Talk:Predictive text

Non-Cellphone Predictive Text
Openoffice.org has a feature that is a kind of predictive text though it only activates for words over a certain length. Should this be included? Samineru 04:08, 27 August 2007 (UTC)

No, Word completion covers that. — Preceding unsigned comment added by 24.31.16.195 (talk) 22:21, 19 November 2011 (UTC)

Merging with T9 page
T9 is a just a brand of predictive text. It should be clearly marked as such in the wikipedia, unless you are going to merge "vacuum cleaner" with "hoover", "record player" with "victrola" etc. I would rather make the title of the t9 page be "t9", not "t9 (predictive text)" as it is presently.

Reply: I agree with what you're saying except it should be considered that Nokia has branded T9 as "Predictive Text" within their phones. As a result, many people know T9 as predictive text (and not as T9).

The people at T9 should complain to Nokia. They are at risk of losing their trademark, see Genericized trademark. Don't think wikipedia should contribute to T9 genericide. Lesismore 04:33, 14 April 2007 (UTC)

Czech t9
I have T9 on my old Nokia with Czech dictionary, and sometimes it seems to offer some prefixes which doesn't correspond to any word in Czech. Does the T9 algorithm use some sort of hashing or dictionary compression? Thanks, Samohyl Jan 09:59, 7 August 2005 (UTC)

T9?
I was wondering where the designation T9 came from...? Erath 23:37, 19 October 2005 (UTC)

T9
T9 stands for "text with 9 keys". Although you'd normally use the "0" and some more to write a text message, the letters, being the most important, are only found on the keys "1" to "9".

As for how the thing works, I don't believe it is done with a dictionary since a dictionary look-up is extremely unreliable in a domain where words are routinely shortened or misspelled for reasons of humour or conciseness. I think it's done with a statistical model of trigramm probabilities, e.g. "from all possible combinations of letters possible when the input is "1-3-3", present those combinations of three letters (trigramms) in a row as suggestions that you have seen in a training corpus of text messages, ordering them by frequency in that corpus". The user can then edit the internal dictionary to augment the mechanism. Here, a simple dictionary lookup is used and is given priority. It might be that unseen trigramms are also permitted as suggestions somewhere further down the list to account for the fact that the training corpus is never complete. The algoritm cannot tell whether they are really rubbish or simply very rare.

Hope it helps, Chris

t9 code in c programming language
does any one have a code for t9 in c programming language .....i would be obliged

RE: T9
Sounds like a certain DS&A assignment we were set.....

Spamming
This article is being constantly spammed by User:Asabir (aka User:212.127.30.42) who adds references to AdapTex. After reverting a previous attempt, I included the reference to this product in the same format as the others listed in the Companies and products section, to be fair (there are other products there, after all). However, I think the entire section should probably be removed. - Gobeirne 10:38, 1 November 2006 (UTC)

I tried to stop Asabir from spamming also, twice. See my edits of Oct 6, and Aug 18. Asabir clearly is an aggressor (3 revert rule), decidedly doesn't take a hint, and should be blocked. if you have the authority. However, I do not agree that the section should be dropped, as many people comming to this page will be interested in actual implementations of predictive text ideas, and there are only a handful of active companies in this space. Adaptex seems to be an actual thing, or is intended to be an actual thing, and could be fairly included in the list of products. Lesismore 15:00, 1 November 2006 (UTC)

Additional reference, broken link
In the Additional reference section, the "T9 training" link leads to a 404 web page error.

If anyone happens to know what this was supposed to point at, or has a similar example, it might be helpful for users who are looking for additional information. Sawta (talk) 22:54, 12 April 2009 (UTC)

Does it work for multiple languages in one sentence?
I can never make T9 working: it always suggest something different from what I want. Could that be because T9 only works for one language in a sentence i.e. no mixing of languages is allowed? Andries (talk) 22:43, 17 October 2010 (UTC)

Major copyediting and reviewing
What started as wikignomeing the lead section, ended as copyediting, rewriting, and reforming many paragraphs in the lead and Background sections. I moved things around between them and added information, as per the edit summaries. I hope any objections to the modifications in there entirety will be considered for discussion here first instead of a complete rejection, or reversion back to to where I started from. &mdash; Cp i r al Cpiral  01:19, 23 March 2011 (UTC)

This article has taken a major step backwards. Notably, real citations have been removed and replaced with (badly written) speculation. Be bold and all that, but please do actual research and supply solid references. There are plenty of scholarly articles on this subject out there. Ipthief (talk)

Out of date?
Phones nowadays don't have multi-character keys, but we still talk about predictive text, do we not? The thing where the phone suggests what word you might want to type next - is that not called predictive text these days? If so, the definition given at the start of this article is badly out of date. W. P. Uzer (talk) 17:54, 4 August 2020 (UTC)

This article is out of date. Predictive text now rarely refers to technologies such as T9. Rather it refers to predicting the partially-typed word or the next word input on a full keyboard. As examples see these pages from Apple and Samsung, which include providing suggestions for the next word. Wikipedia has an article on autocomplete, but the feature in both Android and iOS, where most people encounter it, is referred to as predictive text rather than autocomplete. Seabonn (talk) 16:55, 2 March 2021 (UTC)

Disambiguation page needed
With evolving technology, especially widespread use of smartphones, this article is is largely historical. A disambiguation page may be needed. At least some better ways to advise users that this is a largely historical note, focused on early mobile phones with fixed physical, often numeric, keyboards and that current usage of the term predictive text has changed.

CuriousMarkE (talk) 06:39, 23 April 2023 (UTC)


 * I have added a paragraph early on as a placeholder prior to a more substantial change.
 * Personally, I would like Predictive Search and the modern use of Predictive Text to have their own pages. The ability, or lack of ability, of these technologies to learn from users corrections to poor automated corrections / predictions would also be desirable.  My typing is more error prone over the years.  That may be for a variety of reasons, dexterity, neurological difficulties, bad habits, poor user interface design, unintended consequences of sw changes etc.
 * This one of the few changes, so far,I have to the article, rather than just commenting in the talk pages. There is a big and rather intimidating learning curve to become a more active contributor. CuriousMarkE (talk) 07:04, 23 April 2023 (UTC)
 * My change was reverted by Belbury on 26th October 2023. They believe my paragraph was clumsy and they felt the context was clear enough.  Sadly, I disagree.  As can be seen, I am not a confident, or regular, editor of Wiki pages.  This incident has encouraged me to look more at some of the editing tools to help understand what has happened. I don't think this reversion has been supported by a related change that addresses my concern.
 * I am attempting to make other changes with a similar effect today.
 * Note, I tried to raise this on the Disambiguation Talk page. Sadly, my concern about the general issue wasn't addressed there.  I will consider revisiting the general issue soon as well. CuriousMarkE (talk) 00:59, 13 December 2023 (UTC)