Word Prediction—What's Good Enough | Don Johnston Assistive Technology

Word Prediction—What's Good Enough?

So, you've concluded that word prediction is an effective way to bridge the gap between thoughts and written expression. Now, how do you decide on a word prediction program? What are the distinguishing differences in word prediction programs? And further, what is good prediction worth?

Would you ever compromise the quality of medical attention you seek for your family? Would you buy an automobile that tested "Poor" in consumer safety reports? What about air travel — would you be comfortable traveling in an obviously unsafe airplane? Reading and writing skills are just as critical and we need to consider them as such. Children with learning difficulties only open the window of effort on occasion — when that window opens we must be there with the best tools to ensure that they have a successful experience. We must look at word prediction with critical eyes if we are to offer the best possible opportunity for success.

Over the years we received hundreds of questions about word prediction relating to grammar, topical vocabulary, phonetic spelling and the types of prediction available. We understand there is some confusion about functionality and supports of similarly named features.

History of Co:Writer's Word Prediction

Co:Writer® was produced and is continually updated in collaboration with Paul Schwejda and Judy McDonald, who are true pioneers and have been perfecting word prediction for over 30 years. They began their investigation into word prediction back in the late 70's. In the early 80's they produced "Predict It." but memory on those old Apple II computers caused some limitations. When the Macintosh came to market, Co:Writer® was born. The Macintosh presented the opportunity to develop word prediction with many enhanced features. Since Co:Writer was released in 1992, and technological advances have continued to be made, they have remained singularly devoted to the innovation, refinement and perfection of Co:Writer's prediction and it's ability to support struggling writers.

Three Types of Word Prediction

Word Completion

You'll find this in applications like MS Word and Excel. If you begin to type, for example, the date "j-a-n-u," the application predicts January xx, xxxx (the current date).

Bigram/Trigram Prediction (patterns)

This type of prediction utilizes two and three word patterns, and the frequency in which those two or three words appear together. There are 3 key issues with regard to this type of word prediction:

The effectiveness of the pattern prediction hinges completely on the types of text the developers have used to analyze word patterns in general. Developers have used such common mediums as newspapers. Vocabulary and word patterns found in newspapers do not support the needs of a beginning writer.
This prediction is not effective when new words or topic words are entered because pattern prediction is dependent on the word patterns established by analyzing other writing. This type of prediction only becomes effective after repeated use. The impact on the student is that correct predictions do not occur efficiently because the software needs to monitor the word usage to establish proper prediction. For a struggling writer, if the word is not predicted properly it will not be used as frequently.
New words entered need to contain all forms of the word to be predicted correctly because prediction is not based on grammar and root words. This requires you to enter every form of any given word. For example, if the word explore was not in the dictionary, the word would need to be entered multiple times: explore, explores, explored, exploring, explorer for prediction to occur for all forms of the word. Consequently, a struggling writer would only be presented with that form of the word that was entered.
Pattern prediction is the type of word prediction offered by all producers of word prediction — all except Co:Writer.

Linguistic Word Prediction

The type of prediction Co:Writer uses is called "Linguistic Word Prediction." With Linguistic Word Prediction, Co:Writer knows the grammatical value of each word in its dictionaries. When Co:Writer learns new words collected from articles or student writing, it automatically assign grammar to them. With grammar-based intelligence, Co:Writer can accurately predict words within the framework of valid sentence structures. It also gives flexibility to the words it learns by automatically predicting in multiple tenses and usages.

Dictionary Size

Dictionary size has been somewhat confusing to customers when comparing word prediction programs. One would assume that a dictionary that contains 100,000 words contains more prediction choices than a dictionary like Co:Writer's largest option, which offers only 40,000 words. False. Because Co:Writer uses linguistic prediction, its dictionaries often utilize root words. Grammar-smart Co:Writer understands that words can be used in different capacities, as a noun, as a verb and as an adjective, therein eliminating the necessity of carrying separate entries of the word in the dictionary. In addition to reducing dictionary size, linguistic prediction sharpens word usage, providing superior prediction as students begin to increase productivity and write more descriptive sentences. In addition to the hundreds of thousands of elementary, middle school, and high school students, there are a number of college-level students and even a few doctoral candidates using Co:Writer.

Try This

If you have access to word prediction alternatives you might try the following sentence patterns as a comparative test.

Dogs are furry. (noun, verb, adj.)
(When you typed d-o, did you get plural noun choices? You did using Co:Writer!)
Three very mangy dogs ran down the street. (adj., adv., adj., noun, verb…etc.)
(After typing three, very, mangy, you get plural noun choices, then you get plural verb tenses. This is the type of prediction you will NOT get with pattern prediction products.) All word prediction should be able to get simple sentences like, "I am happy." or, "The fish is swimming." with minimal keystrokes. If a student has to type fis for "fish", the prediction is not good enough. As students learn to write richer sentences, Co:Writer goes beyond 'simple' by utilizing advanced linguistic prediction. It provides the critical modeling of word forms, subject-verb agreement and pronoun and article use.

Critical Features and Performance Standards

Grammar Usage and Dictionaries

There are five main dictionaries to choose from within Co:Writer. A main dictionary is selected based on the writer's ability level and provides appropriately leveled word choices that help the student build good first sentences. In addition to utilizing words from the main dictionaries, Co:Writer automatically collects and assigns grammar to most new and learned words too. For example, if a student is using the Intermediate Dictionary (12K) and wants to write about Mars and the rover "Spirit", which is not found in the main dictionary, Co:Writer will automatically collect, assign grammar and, in correct context, predict the proper noun "Spirit" the next time the student types the word. Typically, new or collected words will only consist of proper nouns and acronyms. When using Co:Writer, no less than three dictionaries are always running simultaneously-a main dictionary, a collected words dictionary AND the writer's personal dictionary. Co:Writer can refine predicted words even further by applying…

Topic Dictionaries

Topic Dictionaries — powerful vocabulary tools-are lists of words grouped together by a specific topic or content area. Rather than laboring over how to spell Pterodactyl or Tyrannosaurus, students can focus on writing for meaning and retelling their knowledge. Applying a Topic Dictionary increases students' efficiency by getting to content-specific words in just one or two keystrokes! There are over 500 Topic Dictionaries for Co:Writer! Additional Topic Dictionaries can be made in less than two minutes using web sites, e-books, textbooks and worksheets. Anything that can be saved as a text file can be made into a Topic Dictionary!

Phonetic Spelling

Many companies that produce word prediction tout phonetic spelling. Unfortunately, some of the most widely-used prediction software only contain a few common phonetic patterns. This leaves you with the job of assessing, and ultimately inputting, the rest-again, not good enough. Co:Writer's FlexSpell™ provides every conceivable letter-pattern students will try in an attempt to spell-out words. A great deal of assessment, using writing samples of students with developmental spelling difficulties, has been done to provide scaffolding for writers who struggle more severely. FlexSpell can be adjusted to work after just one letter is typed, for example if a student types the letter "u" Co:Writer will predict the word "you", or FlexSpell can be set to provide phonetic spellings only after two or three letters have been typed. FlexSpell was designed to remove the mechanical barriers that keep many students from expressing themselves and their knowledge through writing.

Try This

The blk jragon flu ovr the gra lfnt. (The black dragon flew over the gray elephant.)
R u hpy to ce me? (Are you happy to see me?—only 6 keystrokes!)
I no hw to nsr the fone. (I know how to answer the phone—less than 8 keystrokes)

Note—You will not need to type in all of the letters given above, but they are included as a point of reference. Try these types of spellings or, better yet, use your students' writing samples to compare the prediction for yourself!

eWord Bank

Available only in Co:Writer, the eWord Bank automatically presents a bank of topic-specific words when a Topic Dictionary is selected. This is especially useful for students who stare at the blank screen — unable to translate their thoughts into words. Quickly scanning through the topic words helps students brainstorm and formulate their thoughts. After brainstorming, the eWord Bank helps build core vocabulary by turning your students' receptive vocabulary (the words they "know" but don't necessarily use in their writing) into productive vocabulary used in writing.

Set Up Time and Customization

Most word prediction programs are NOT plug and go. However, with Co:Writer 6 students can set themselves up in less than a minute. With FlexSpell and the new Easy Create Topic Dictionary functionality students can write on the topics they choose with a simple cut and paste from any webpage! The student now drives much of the time intensive work most educators struggled with. For the savvy teacher who wants to further tailor the settings, there are a number of options and preferences that can fine-tuned. The Getting Started section of Co:Writer's manual is a step-by-step tutorial that can be used as a quick reference or as a training resource. Additionally, product training, customer and technical support are readily available.