




































Interactive web-based learning of corpus-generated phrases 
Dougal Graham & Christopher Osment 

AsiaCALL Online Journal 

Vol. 9 (2014) | http://asiacall.info/acoj | ISSN 1936-9859 A1 

Interactive web-based learning of corpus-generated phrases 

Dougal Graham 

Christopher Osment 
King Mongkut’s University of Technology Thonburi, Thailand 

Corresponding author: dougal.gra@kmutt.ac.th  

 
Submitted 30 January, 2014; accepted in final form 15 May, 2014 

 

 
 

Abstract 

This article will review the development of a partially-automated online teaching system developed 

to help intermediate-level engineering students learn appropriate English phrases in an English as 

a Foreign Language (EFL) environment. The corpus contains approximately 1.2 million words. 

Rather than focusing solely on keywords, these materials highlight useful collocations, and genre -

specific uses of common words and phrases. The online interactive materials are able to generate 

a relatively large number of unique exercises for students. These exercise allow students to not 

only practice content that they have previously learned, but in fact to learn new material that they 

have not previously studied. In this way, the materials can be used either by students alone, or i n 

conjunction with in-class use of the paper-based materials. 

 

Keywords: Corpus, English for Specific Purposes, Online, Data-Driven Learning, Formulaic 

Language 

 

 

Introduction 

A large body of research has demonstrated the importance of formulaic language (collocations, formulaic 

phrases, multi-word units, lexical bundles, etc.) in the production and comprehension of fluent natural 

language. Schmitt (2010, pp. 117–120, 142) and others (Nattinger & DeCarrico, 2009) explain that 

formulaic language is important for conveying routine meanings, lexicalising various functions, social 

functions, discourse organization, and precise information transfer as well as promoting fluency and Biber 

(2007; 2004)’s lexical bundles focus on the use of formulaic language for discourse and pragmatic 

functions. Formulaic language also brings cognitive processing benefits as it is easier and faster to process 

and generate once learned than non-formulaic expressions and increases retention of content (Tremblay, 

Derwing, Libben, & Westbury, 2011) compared to incorrect use of formulaic language which increases 

processing demands (Millar, 2011). 

 

Formulaic language, then, has clear benefits for English as a Second or Foreign Language learners. If a 

learner can master even some formulaic language, it will greatly increase their ability to both produce and 

comprehend language quickly and easily, resulting in greater fluency, comprehension and 

comprehensibility. This language is also useful for learners in an English for Specific Purposes (ESP) 

context. Formulaic language differs between registers (Biber & Barbieri, 2007) and contexts (Conrad & 

Biber, 2004) and knowing the correct formulaic language to use within a given community can not only 

help a speaker communicate clearly and effectively within the group, but also help the learner mark 

themselves as a member of the language community or show social solidarity (Norbert Schmitt, 2010, p. 

10) increasing trust and social cohesion. 

 

When learning formulaic language, frequency of exposure, and frequency of use are the two most 

important factors in determining whether a learner will acquire a given piece of formulaic language or not 

(Ellis et al., 2008; Huang, Wible, & Ko, 2012). The result of the need for both repeated exposure and 

repeated use is then a focus on repetition and rote-learning style activities for the study of formulaic 

mailto:dougal.gra@kmutt.ac.th
mailto:dougal.gra@kmutt.ac.th


Interactive web-based learning of corpus-generated phrases 
Dougal Graham & Christopher Osment 

AsiaCALL Online Journal 

Vol. 9 (2014) | http://asiacall.info/acoj | ISSN 1936-9859 A2 

language. For example, Yu (2009) recommends explicit memorization as a method by which to improve 

student performance for formulaic language. Similarly Lewis (1997, p. 51) explains that many researchers 

recommend “meeting” a word at least seven distinct times in order to learn it properly, and one can 

reasonably imagine extending this concept to formulaic language. While such an approach may be useful 

for learners, it does have certain drawbacks. 

 

Some of the drawbacks of a memorization, or rote-learning approach are time-consumption (Swan, 2006), 

student demotivation and disengagement (Bowen & Marks, 1994, p. 101), and a focus on surface-level 

memorization, rather than understanding (Biggs & Moore, 1993, p. 215). Bowen (1994) states that 

students often feel that memorization and rote learning tasks make the learning seem like a “difficult , even 

painful process”. Such demotivation could lead to students who do not try, or feel that the effort to learn 

the language is not worthwhile. Furthermore, focus only on the surface-level form of the structures, while 

ignoring their context and usage makes the target language more difficult to learn by ignoring valuable 

information about when the language is appropriate to use. 

 

An alternative to rote-learning is to take a data-driven learning (DDL) approach. DDL is an approach to 

language learning that focuses on using genuine language examples (data) as the core from which a 

student is then guided to “learn how to learn” the language (Johns, 1991, p. 31) cited in (Boulton, 2011). 

Because of this focus on the student’s ability to learn for themselves, the direct access to the data, and the 

focus on induction (Johns, 1991, p. 29, as cited in Boulton, 2011), DDL can be considered to fall within 

the category of a student-centered approach to learning. A data-driven learning approach could be used to 

provide a student with the target language they need and are immediately interested in learning, couched 

within a variety of real contexts. This would then allow the student to learn and discover the formulaic 

language inductively and within its appropriate context, rather than in a decontextualized rote manner. 

 

This paper will present an alternative automated approach to the teaching of formulaic language in an ESP 

setting which avoids the problems associated with rote learning. This project makes up the Corpus-based 

Engineering English Materials (CEEM) website which is available at (http://crs2.kmutt.ac.th/ceem). We 

will first outline our general theoretical principles for the creation of the CEEM materials following which 

we will give a brief overview of some currently available materials, before finally giving a detailed 

discussion of our approach to the generation of materials. 

 

Principles 

In an attempt to achieve the necessary repetition for the learning of formulaic language without resorting 

to rote learning, and keeping in line with the principles of DDL, we have determined five key principles to 

use in the development of our materials whose rationale we will discuss in more depth below. The first 

two principles are a direct attempt to avoid the issues with rote learning while the final 3 principles are 

more closely associated with DDL and a student-centered approach. The first principle is to avoid pure 

memorization, and the second to employ variation in the materials presented to students. This will help to 

prevent rote memorization as the exact context of presentation will be constantly varied and changing. 

The final three principles which fall under the rubric of a student-centered DDL approach are an attempt 

to foster student independence, the use of authentic language, and the fostering of self-motivation in the 

students. The goal of the third and fourth principles are to foster independence by creating materials 

usable either in or outside of the class with or without teacher support, and by presenting authentic 

language, to force the student away from a rote focus on the formulaic language unit as an entity separate 

from context and to instead see it within the context of the various texts within which it appears. The goal 

of the final principle is find ways to encourage students to take charge of their own learning and not be 

entirely dependent on the teacher or classroom as motivation for learning. 

http://crs2.kmutt.ac.th/ceem


Interactive web-based learning of corpus-generated phrases 
Dougal Graham & Christopher Osment 

AsiaCALL Online Journal 

Vol. 9 (2014) | http://asiacall.info/acoj | ISSN 1936-9859 A3 

 

Independence  

In order to allow students greater freedom and self-direction, it is necessary that students be able to access 

the learning materials either in or out of class. That is to say they should encourage student independence 

and work equally well with or without a teacher's support. This implies the need for materials to be freely 

available online, and formatted such that they are equally useful and interesting on a personal computer, a 

notebook, a tablet, or a small smart phone. In this way, the materials can be accessed either outside of 

class, or within a classroom setting with additional teacher support. 

 

Secondly, in order to foster true independent student-centred learning there must be choices available for 

students. We have taken this to imply the need for both a choice of multiple types of activities to suit a 

variety of learning goals and styles, as well as choice within each activity for the content that can be 

accessed. 

 

Finally, we understand that not all students will be ready or able to immediately use these exercises. For 

that reason, we have developed complementary introductory exercises that teachers can employ in class. 

These exercises serve as both awareness-raising opportunities and a chance to practice useful language. 

Once students have taken part in these activities with a teacher’s guidance they will be more ready to 

reinforce their new knowledge by independently taking part in the online activities which we have 

created. 

 

Avoid Pure Memorization 

While frequent exposure is clearly necessary for the development of a powerful and useful vocabulary of 

formulaic language as discussed above, it does not necessarily entail that pure memorization is either 

sufficient for or necessary to the acquisition of formulaic language. An alternative approach is to ensure 

that the learner is actively engaged in the language, not only in reading it and looking for their own 

patterns and rules, but taking part in interactive activities which help to highlight important information 

and require active input and consideration from the user. These exercises can be used in conjunction with 

scaffolding to promote inductive learning of formulaic language, lexico-grammatical profiles or other 

concepts, and to ensure that students will be able to use these skills independently in the future. 

 

Employ Variation 

While variation helps to avoid pure memorization by exposing students to a variety of different contexts 

and situations for use, it also addresses the problem of over-structured exercises that occurs with rote 

learning. We propose that a useful source of exercises should produce a large number of different 

exercises for students with a low probability of a specific set of exercises being repeated. With most 

existing exercises, both online and paper-based materials, it is extremely difficult to avoid the explicit 

repetition of exercises and examples. Some online activities such as FLAX (Witten et al., 2013) which 

promotes language exploration, do employ significant amounts of variation by dynamically tapping into 

corpora. However, these activities do not actively help to foster a student's ability to recognize linguistic 

patterns, or to put into use the linguistic patterns that they may be able to recognize. Moreover, the 

abundance of information that many corpora possess can be daunting to learners who are unfamiliar with 

using corpora for language learning. 

 

Motivation 

Finally, none of the above principles will be useful if students are uninterested in the activities presented. 

To address this issue, we believe that the activities and exercises should fall on a continuum from fun to 

challenging, and should incorporate aspects of gamification or educational games. Gamification 



Interactive web-based learning of corpus-generated phrases 
Dougal Graham & Christopher Osment 

AsiaCALL Online Journal 

Vol. 9 (2014) | http://asiacall.info/acoj | ISSN 1936-9859 A4 

(Deterding et al., 2011) is the introduction of game-like aspects to what is traditionally not considered a 

game situation. This can be the introduction of scoring, comparative rankings, or other systems commonly 

used within the realm of games. These systems can, if used correctly, encourage repeated use, and provide 

a more relaxed approach to the use of the materials. The simplest and most useful type of gamification is 

the use of instantaneous feedback to let the user know how they are progressing, and/or to score them. 

This fits very well with the need to provide useful scaffolding to students as they progress through an 

activity. Currently, the project only supports very loose gamification in the sense that it contains both 

educational games and a partially gamified set of exercises without any of the long-term scoring 

accumulation, heuristics, or directly competitive aspects of fully gamified applications. 

 

Authentic Language 

The use of authentic useful language is a key principle of DDL so it is necessary here to discuss our target 

audience, corpus, and general learning goals for the CEEM project. This project is intended to serve the 

interests of first or second-year undergraduate students at engineering universities, especially in 

developing countries. The work is based upon a corpus of first-year English textbooks from a Thai 

Engineering University's “International Programme”. Work by several scholars (Evans & Green, 2007; 

Nurweni & Read, 1999) has found that students at such institutions often lack critical linguistic ability 

necessary for their academic studies, which are frequently entirely in English. Students have great 

difficulty in comprehending both their textbooks and their exam questions, and given the importance of 

formulaic language, the teaching of formulaic language seems an appropriate starting point to address the 

issue of poor comprehension of basic instructional materials. 

 

In order to create exercises using meaningful in-context examples of engineering English that students 

could expect to encounter in their academic studies, the Engineering English Corpus (EEC) was created 

using, on average, 45,000 word representative samples from each of the textbooks that the students would 

be using during their course of study in the international programme. These samples included sections 

from each chapter and from a variety of styles of writing, from explanations to practice exercises and 

questions. The books were digitized, and Adobe Acrobat's optical character recognition software was used 

to translate the files to text format. The final corpus comprises 29 textbooks and approximately 1.15 

million words of text (See appendices for full details of corpus composition). 

 

There are many theoretical approaches to formulaic language. We have not discussed them in much detail 

as they are not the primary focus of this paper, but they do bear mentioning in order to make clear what 

our learning goals were in the development of the CEEM Project. Approaches to formulaic language 

range from solely frequency based lists, such as Biber's (2007; 2004) lexical bundles, to general formulaic 

language described in the Phrasal Expressions List (Martinez & Schmitt, 2012), comparative EAP studies 

such as the Academic Formulas List (Simpson-Vlach & Ellis, 2010), to gapped phrase approaches such as 

phrase-frames (Fletcher, 2011), or concgrams (Cheng, Greaves, & Warren, 2006). For the purpose of this 

work, we have focused on high-frequency four-word sequences, which either do not frequently occur in 

other contexts or, in the context of Engineering English, have different lexico-grammatical profiles. We 

have taken a partially intuitive approach, to this task. This approach will be outlined in detail below. 

 

Currently available materials  

Generally, there are two types of DDL materials that are currently available online : “discovery materials” 

or “testing materials”. Discovery materials are almost completely unstructured and are designed only to 

facilitate a learner's ability to locate useful corpus data with which to inform their language use. For 

example, FLAX and the Compleat Lexical Tutor (Cobb, 2013) focus primarily on allowing students to 

access concordance lines that might be useful to them, either by writing and searching for a word about 



Interactive web-based learning of corpus-generated phrases 
Dougal Graham & Christopher Osment 

AsiaCALL Online Journal 

Vol. 9 (2014) | http://asiacall.info/acoj | ISSN 1936-9859 A5 

which they are unsure, or by choosing words from a word list. Such materials are useful to advanced 

learners who know what they want to study and learn, but they may not provide enough guidance for 

lower level learners. 

 

Testing materials, on the other hand, test a student's knowledge and provide instant feedback, at least 

partially meeting our criteria for motivation. Ideally, they also incorporate some sort of scaffolding to 

enable learning throughout the testing process. The Lexical Tutor site contains a set of exercises that test 

collocational knowledge comprising approximately 30 items. The collocation tests are interesting in that 

students are provided with scaffolding in the form of concordance lines from which to induce a correct 

answer. Such materials then, meet our requirements for authentic language, self-motivation, and 

independence, but they are unchanging and do not contain sufficient variation for a learner to be able to 

practice difficult language on more than one occasion. The powerful FLAX tool allows teachers to 

quickly and easily generate activities for students from given readings and articles, however, as a tool 

which works on a single text, it is unclear whether these activities will lead to useful generalizable 

language rules for students. 

 

This pattern of many small sets of materials which are useful, but not fully meeting our requirements 

demonstrates the need to look for easier ways of creating reusable and automatically variable materials, 

rather than attempting to create a large number of static non-reusable exercises. 

 

Creation of Materials 

Using the EEC described above, the Engineering Word List (EWL) was created. The EWL, which can be 

accessed online at http://crs2.kmutt.ac.th/ceem/ewl was created using a combination of an empirical and 

intuitive approach. First, a raw word list was created using the AntConc concordancing application 

(Anthony, 2013). The list was restricted to words occurring in at least 10 textbooks, or in one of the core 

textbooks (physics or calculus) used by all departments for their engineering students. This list was sorted 

from highest to lowest frequency words, and was then triaged by removing non-content words including 

determiners, prepositions, conjunctions, pronouns, and single-character symbols used in various 

mathematical formulae. For a sample comparison of the ten most frequent words in the corpus as a whole 

compared with the ten most frequent words in the EWL, please see Table 1 below. 

 

Table 1: Ten most common words in the EEC and the EWL 

Rank in EEC EEC Raw Frequency  Rank in EEC EWL 

1 the  5 is 

2 of  10 are 

3 a  11 be 

4 and  34 have 

5 is  41 used 

6 to  45 figure 

7 in  48 system 

8 that  52 force 

9 for  54 use 

10 are  56 energy 

 

Following the triage, the words were organized into 418 word family groups, containing the top 1,416 

content words. This list was examined intuitively in order to locate 12 highly frequent words showing 

http://crs2.kmutt.ac.th/ceem/ewl


Interactive web-based learning of corpus-generated phrases 
Dougal Graham & Christopher Osment 

AsiaCALL Online Journal 

Vol. 9 (2014) | http://asiacall.info/acoj | ISSN 1936-9859 A6 

significant variation from standard usage to use a starting point for the creation of exercises with the goal 

of eventually producing exercises for all the words on the list. These words are: applied, based, common, 

consider, defined, difference, done, equal, following, given, have, and how. For each word, two or three of 

the most common 4-word MWUs (phrases) containing that word were selected from the corpus as 

exemplifying engineering-specific language that was likely to cause difficulties for the students.  

 

Phrases were selected not on the basis of non-standard lexis, but on the basis of non-standard lexico-

grammatical profiles. A phrase such as “a bipartite graph” is relatively common in the data (LL 53.24 vs. 

BNC), however it exemplifies a technical term which will certainly be covered in the course of their 

regular lectures. The interesting and useful phrases for the students will be ones that contain everyday 

common words, but that use them in ways that the students are not familiar with such as “is defined to 

be”. Students are all generally familiar with the individual words in this phrase, but unfamiliar with the 

use of them together in this ordering and to achieve this function. These phrases, and their genuine 

contexts from the corpus were used to generate random DDL exercises for the students. 

 

Explanation of CEEM Materials  

This section will contain an in-depth description of the goals for each of the four types of exercises that 

are currently available from the CEEM project website including materials available for teachers' use. 

First we will discuss the in-class activities, then we will describe the three current on-line activities: sight 

words, hangman, and the phrase challenge. It is important to keep in mind that these materials are 

intended to be holistic and incremental. While the primary activity does in fact take into account all the 

principles that we have discussed above, some of the introductory exercises do not. These issues and the 

reasons for our decisions will be discussed below. 

 

In-class exercises 

The printable in-class exercises (available at http://crs2.kmutt.ac.th/ceem/teacher/exercises) are based on 

the 1,000 most frequent vocabulary from the EWL. As such, they adhere to our principle of using genuine 

language in that they are words which students will definitely encounter frequently throughout their 

studies, and in addition, the exercises use genuine sentence-length excerpts from their texts. These 

exercises have three main pedagogic goals which are to teach collocations, to contrast apparent synonyms, 

and to raise learners' awareness of these concepts and of skills necessary to make these inferences on their 

own from natural language. 

 

It is necessary to clarify the place of these exercises within the context of CEEM. These exercises, as 

classroom-based, printable exercises are neither highly varied (and are clearly single-use), nor particularly 

student-centered. However, the purpose of these exercises is to raise a student’s awareness of corpus-

based approaches to learning, and to provide them a chance to learn basic skills for inducing usage from 

real-world examples. These exercises are particularly intended for lower level learners who may not have 

as much language learning experience as a more advanced student. 

 

These exercises take collocations into consideration by trying to asking students to utilize noticing so that 

students will become aware of common collocations in engineering writing. We begin with simple 

collocations but also note that some of them can build up into a longer multi-word units. 

 

As well, there are exercises that contrast apparent synonyms. Students are often unaware of how apparent 

synonyms frequently have significantly different usage patterns. These types of exercises are meant to 

foster an understanding in students that even though two words may have synonymous meanings, the 

usage of such words can be quite different. Comparisons of apparent synonyms show that their actual 

http://crs2.kmutt.ac.th/ceem/teacher/exercises


Interactive web-based learning of corpus-generated phrases 
Dougal Graham & Christopher Osment 

AsiaCALL Online Journal 

Vol. 9 (2014) | http://asiacall.info/acoj | ISSN 1936-9859 A7 

usage is quite different, and these exercises help students to see those differences. For example, “make” 

and “create” appear to many students to be completely synonymous, however, in our data “create” is often 

used when discussing data modelling or mathematics as in the sentence fragment “…and the capability to 

create complex geometries…” while “make” is more often used with the creation of a physical entity or to 

communicate how to do something as in “to make your drawings clear and easy to read”.  

 

Awareness raising is an important element in our exercises. These exercises raise students' awareness of 

the above concepts and help them to develop the tools to learn more efficiently in future study. They give 

the students access to a variety of examples from real usage and try to ask questions which will lead the 

students to draw inferences about the words and how they are used in natural language.  

 

On-line Exercises: Sight Words 

The “sight words” activity is an online low-demand game-like reading practice exercise. While the 

primary focus of the CEEM activities is the teaching of phrases, it is important for students to be aware of 

the highly frequent words in the language they are reading, and then to make the connection between 

those words, and their usage in phrases. By given students a chance to focus only on words, before seeing 

those words in the context of phrases, we hope that we will help to raise a student’s awareness that words 

are not used in isolation, and to build incrementally from a small simple activity, to a more complex one, 

as will be described below. 

 

In this activity, students choose between three and six words, which they would like to practice. These 

words have been chosen from among the most frequent words in the EWL. This activity attempts to 

develop automaticity (Logan, 1985) in student reading to improve reading speed and proficiency. 

Proficient readers are able to distinguish words more quickly and efficiently than beginner readers who 

must explicitly sound out the spelling in their heads as they go (Biggs & Moore, 1993, p. 341). The goal 

of this activity is three-fold: firstly, it aims to increase student awareness of some of the most frequent 

words in the corpus, secondly it aims to help increase reading speed by helping students develop 

automaticity in recognition of the target words, and thirdly it is meant to be a fun and engaging activity to 

help students become engaged with the material. 

 

Student autonomy is promoted by asking students to choose which words they would like to practice. 

They are also given the option to randomly select a set of words. As can be seen in Figure 1, the activity 

presents students with a countdown timer incremented by two and a half seconds for each word. A target 

word is shown in the top center of the screen above the timer. Upon clicking the “start” button, students 

are presented with a grid containing six buttons. One button contains the target word, and the 5 other 

contain similarly spelled words and non-words. The student must select the correct word from the grid as 

quickly as possible. The timer is not fixed per word, but per game session, so that if one word takes 

slightly longer, but another less time, the student will still be able to complete the activity in the time 

allotted. When a potential answer is selected, the countdown timer stops, and the student is presented with 

feedback displaying the correct answer if their answer was incorrect (see Figure 1, below). 

  



Interactive web-based learning of corpus-generated phrases 
Dougal Graham & Christopher Osment 

AsiaCALL Online Journal 

Vol. 9 (2014) | http://asiacall.info/acoj | ISSN 1936-9859 A8 

 

 

 

 

 

 

 

 

 

Figure 1: Sight words activity with incorrect word darkened and correct word lightened 

 

As students progress through the activity, a coloured wheel in the top-left hand side of the screen displays 

their score, and at the end of the activity they are given a full feedback report detailing which words took 

the longest for them to find, and which they answered incorrectly allowing them to note those words need 

further practice. 

  

This activity meets our general criteria described in the first section in that it is classroom-agnostic 

(online), it is not focused on pure memorization, but rather on exposure, it fosters self-motivation through 

immediate direct feedback and scoring, and it employs variation in both presentation and content. The 

variation in this activity warrants some brief clarification. Firstly, the six tiles are randomized, to ensure 

that the presentation is not predictable. Secondly, the list of five similarly spelled words is selected at 

random from the top twelve similarly spelled words  for a given target word, ensuring that the student will 

be presented with a novel experience each time the game is played. 

 

On-line Exercises: Hangman 

Hangman is a classic educational tool used to practice spelling, create an environment for repeated 

exposure and to foster motivation through game-like aspects. In the CEEM Project website, we use this 

familiar activity not only for the purposes just mentioned but also to raise students' awareness of 

formulaic language sequences as single units. Rather than presenting words for students to spell, they are 

presented with four word phrases. As discussed earlier, these phrases were chosen because they contain 

the core EWL words that we are taking as our focal points for the activities. These phrases then also allow 

the students to see that these words (which they have practised in the Sight Words activity) do not occur 

on their own, but as part of larger units. This again reinforces those words through another exposure 

vector, this time with greater context. Again, this is a fun activity, which, although it is unscored, certainly 

contains game-like elements to increase student interest. 

 

On-line Exercises: Phrase Challenge 

The final activity developed for students is significantly more challenging than the two previous activities 

described. As with the sight words activity, students are prompted to choose words that they would like to 

learn more about; however, in this case, 2-3 phrases are selected that contain each word. The students are 

then presented with a series of gapped sentences on the left side of their screen and a list of phrases on the 

right as shown in Figure 2. 



Interactive web-based learning of corpus-generated phrases 
Dougal Graham & Christopher Osment 

AsiaCALL Online Journal 

Vol. 9 (2014) | http://asiacall.info/acoj | ISSN 1936-9859 A9 

 
Figure 2: Phrase Challenge activity with gapped sentences and possible answers  

 

 

The student then has a maximum of three attempts to match the phrases with the sentences via drag and 

drop. After each of the first two attempts, error correction scaffolding is presented to the student alongside 

their errors. For the first failed attempt, the student will be presented with three sample sentences 

containing each phrase (Figure 3). The student is encouraged to try to find a pattern in the examples. 

 

 
Figure 3: First round of scaffolding in the Phrase Challenge: Example Sentences  

 

 

If, after a second attempt, the student is still experiencing difficulty, an example partial lexico-

grammatical profile is presented to the student (Figure 4). The student is also presented with 3 more 

example sentences to examine. 

  



Interactive web-based learning of corpus-generated phrases 
Dougal Graham & Christopher Osment 

AsiaCALL Online Journal 

Vol. 9 (2014) | http://asiacall.info/acoj | ISSN 1936-9859 A10 

 
Figure 4: Second round of scaffolding with a “pattern” for the phrase  

 

This scaffolding is intended to serve two purposes. In the first example, it gives context with which 

learners can begin to attempt to induce a pattern of usage (lexico-grammatical profile) for the phrase in 

question. However, since some students may not be aware of what a lexico-grammatical profile is, or how 

to find one, the second type of scaffolding is provided to raise learner awareness of some possible types of 

patterns, which they can identify for the phrases. 

 

In order to promote variation, the gapped sentences are randomly selected from the database. This 

combined with user selection of words and the number of sentences stored for each phrase and a 

randomized sentence ordering provides an astonishing number of possible exercises for students. There 

are 12 words per section, of which 3 are chosen to generate phrases, resulting in 220 different 

combinations of phrases that one can study per section. Each word is associated with from 1-4 useful 

phrases averaging ~2.16 phrases per word. Within each of the 220 combinations of approximately 7 

phrases, each phrase is randomly associated with 2 sentences resulting in ~1.9 x 1014 possible exercises. 

Of course, a student would likely notice some repetition well before attempting use of the exercises that 

many times, but it should be sufficient for 3-5 attempts distributed over some time, especially if the 

student varies their target words. 

 

Conclusion 

The teaching and studying of formulaic language is an important consideration in language teaching. The 

findings of corpora from the last twenty years have brought its existence in a variety of genres clearly into 

view. Educators need to address the teaching of formulaic language. Simply teaching lexical lists based on 

frequency counts is not enough.  

 

Moreover, with the recognition that learning does not necessarily require a classroom, especially with 

language, the idea of making materials easily available to learners online is vital. This is particularly true 

for younger learners who have grown up along with the internet and for mature learners who may not 

have time to attend traditional classes, but retain a need and desire to learn a language.  

 

Despite the development of improved and versatile features on the internet, many online learning 

materials still tend to lack interactivity and randomness. If we as educators wish to truly motivate and 

engage our students, then we needed to design language materials that harness this interactivity and 

randomness. Furthermore, the usage of randomness better models the actual variety that students 

encounter when they read. 

 



Interactive web-based learning of corpus-generated phrases 
Dougal Graham & Christopher Osment 

AsiaCALL Online Journal 

Vol. 9 (2014) | http://asiacall.info/acoj | ISSN 1936-9859 A11 

References 

Anthony, L. (2013). AntConc (Version 3.3.5w). Tokyo, Japan: Waseda University. Retrieved from 

http://www.antlab.sci.waseda.ac.jp/ 

Biber, D., & Barbieri, F. (2007). Lexical bundles in university spoken and written registers. English for 

Specific Purposes, 26(3), 263–286. doi:10.1016/j.esp.2006.08.003. 

Biggs, J., B., & Moore, P., J. (1993). The Process of Learning (Third Edition.). Prentice Hall of Australia. 

Boulton, A. (2011). Data-driven learning: the perpetual enigma. Explorations across Languages and 

Corpora, 563–580. 

Bowen, T., & Marks, J. (1994). Inside Teaching. Heinemann. 

Cheng, W., Greaves, C., & Warren, M. (2006). From n-gram to skipgram to concgram. International 

Journal of Corpus Linguistics, 11(4), 411–433. 

Cobb, T. (2013). Compleat Lexical Tutor (Version 6.2). Retrieved from http://www.lextutor.ca/ 

Conrad, S., & Biber, D. (2004). The Frequency and Use of Lexical Bundles in Conversation and 

Academic Prose. Lexicographica, 20, 56–71. 

Deterding, S., Dixon, D., Khaled, R., & Nacke, L. (2011). From game design elements to gamefulness: 

defining gamification. In Proceedings of the 15th International Academic MindTrek Conference: 

Envisioning Future Media Environments (pp. 9–15). Retrieved from 

http://dl.acm.org/citation.cfm?id=2181040 

Ellis, N. C., Simpson-Vlach, R., & Maynard, C. (2008). Formulaic Language in Native and Second-

Language Speakers: Psycholinguistics, Corpus Linguistics and TESOL. TESOL Quarterly, 42(3), 

375–396. doi:10.1002/j.1545-7249.2008.tb00137.x 

Evans, S., & Green, C. (2007). Why EAP is necessary: A survey of Hong Kong tertiary students. English 

for Academic Purposes, 6(1), 3–17. 

Fletcher, W., H. (2011). Phrases in English. Retrieved from http://phrasesinenglish.org/ 

Huang, P.-Y., Wible, D., & Ko, H.-W. (2012). Frequency Effects and Transitional Probabilities in L1 and 

L2 Speakers’ Processing of Multiword Expressions. In S. T. Gries & D. Divjak (Eds.), Frequency 

Effects in Language Learning and Processing (pp. 145–176). de Gruyter Mouton. 

Johns, T. (1991). From printout to handout: grammar and vocabulary teaching in the context of data-

driven learning. In T. Johns & P. King (Eds.), Classroom Concordancing (pp. 27–45). English 

Language Research Journal (as cited in Boulton, 2011). 

Lewis, M. (1997). Implementing the Lexical Approach: Putting Theory into Practice. 

Logan, G. D. (1985). Skill and automaticity: Relations, implications, and future directions. Canadian 

Journal of Psychology/Revue Canadienne de Psychologie, 39(2), 367. 

Martinez, R., & Schmitt, N. (2012). A Phrasal Expressions List. Applied Linguistics, 33(3), 299–320. 

Millar, N. (2011). The Processing of Malformed Formulaic Language. Applied Linguistics, 32(2), 129–

148. 

Nattinger, J., R., & DeCarrico, J., S. (2009). Lexical Phrases and Language Teaching. Oxford University 

Press. 

Norbert Schmitt. (2010). Researching Vocabulary: A Vocabulary Research Manual. Palgrave MacMillan. 

Nurweni, A., & Read, J. (1999). The English language knowledge of Indonesian university students. 

English for Specific Purposes, 18(2), 161–175. 

Simpson-Vlach, R., & Ellis, N. C. (2010). An Academic Formulas List: New Methods in Phraseology 

Research. Applied Linguistics, 31(4), 487–512. doi:10.1093/applin/amp058 

Swan, M. (2006). Chunks in the Classroom: Let’s not go Overboard. The Teacher Trainer Journal, 20(3), 

5–6. 

Tremblay, A., Derwing, B., Libben, G., & Westbury, C. (2011). Processing Advantages of Lexical 

Bundles: Evidence from Self-paced Reading and Sentence Recall Tasks. Language Learning, 61(2), 

569–613. 



Interactive web-based learning of corpus-generated phrases 
Dougal Graham & Christopher Osment 

AsiaCALL Online Journal 

Vol. 9 (2014) | http://asiacall.info/acoj | ISSN 1936-9859 A12 

Witten, I., Brine, J., Finch, L., Franken, M., Johnson, M., Te Taka, Wu, S, et al. (2013). FLAX: 

Interactive Language Learning (Version 2.6). University of Waikato. Retrieved from 

http://flax.nzdl.org/ 

Yu, X. (2009). A formal criterion for identifying lexical phrases: Implication from a classroom 

experiment. System, 37(4), 689–699. 

  



Interactive web-based learning of corpus-generated phrases 
Dougal Graham & Christopher Osment 

AsiaCALL Online Journal 

Vol. 9 (2014) | http://asiacall.info/acoj | ISSN 1936-9859 A13 

Appendices 

 

Appendix 1: Disciplines included in the CEEM Corpus 

 

 

Appendix 2: Textbooks of the CEEM corpus by subject and number of words included 

 Textbook Subject # of Words 

1. Biology 42,857 

2. C++ 50,103 

3. Calculus 59,326 

4. Chemical engineering 46,509 

5. Chemistry 45,350 

6. Database 52,811 

7. Data structure 35,789 

8. Discrete mathematics 50,991 

9. Circuits and circuit analysis 34,585 

10. Engineering materials 53,426 

11. Engineering programming 29,165 

12. Environmental pollution 34,235 

13. Environmental engineering 40,861 

14. Fluid mechanics 39,138 

15. Hydraulic fluids 42,174 

16. Java 28,049 

17. Manufacturing processes 61,837 

18. Material and energy balance 21,950 

19. Mechanical solids 26,501 

20. Physics 88,978 

21. Statics and dynamics 50,302 

22. Statics 36,888 

23. Structural analysis 36,826 

24. Surveying 48,353 

25. Technical drawing 69,228 

26. Thermodynamics 54,149 

27. Wastewater management 24,144 

 Total 1,204,525 

 

Disciplines Included in EEC 

Civil Engineering 

Mechatronics 

Mechanical Engineering 

Computer Engineering 

Chemical Engineering 

Environmental Engineering 

Electrical Engineering 

Materials Engineering 

Production Engineering 

Tool Engineering 

Control Systems and Instrumentation 

Electronics and Telecommunication 


