Wikipedia:Articles for deletion/LetterWise


 * The following discussion is an archived debate of the proposed deletion of the article below. Please do not modify it. Subsequent comments should be made on the appropriate discussion page (such as the article's talk page or in a deletion review).  No further edits should be made to this page.

The result was keep. Randykitty (talk) 15:00, 2 June 2019 (UTC)

LetterWise

 * – ( View AfD View log  Stats )

Complete WP:GNG fail. All the sources are either primary, or, if secondary, not independent, and I can't find any reliable sources that discuss it independently. SportingFlyer  T · C  12:35, 12 May 2019 (UTC)
 * Note: This discussion has been included in the list of Computing-related deletion discussions. SportingFlyer  T · C  12:35, 12 May 2019 (UTC)
 * Note: This discussion has been included in the list of Software-related deletion discussions. SportingFlyer  T · C  12:35, 12 May 2019 (UTC)


 * Weak keep. I considered AfDing this myself, but then I found a few mentions in academic works. Such as the one I've added a few days ago. Can the nom explain how it is a primary or not independent? Well, I admit it is likely from the people who patented it, but a primary academic work is I think generally seen as somewhat reliable? Through it's a conference, not a paper, but in computer science, those are relatively common.--Piotr Konieczny aka Prokonsul Piotrus&#124; reply here 15:49, 12 May 2019 (UTC)
 * When I nominated, the article consisted of a single academic paper written by the authors of the software (primary, not independent - I've seen a few academic papers written by the academics who wrote the software recently, and that does not convey notability on the software they wrote), the software patent (primary, not independent), and three links to the company that wrote the software's web site. That paper has been cited a few times, but considering this is commercial software and not say an academic profile under WP:NPROF, I'm not going to withdraw the nomination. SportingFlyer  T · C  00:33, 13 May 2019 (UTC)
 * It perhaps should be noted on record that my best understanding is Scott MacKenzie the first named on the paper was a prolific writer on human-computer interaction and as far as I am aware was not not an employee of the company (but there may have been a relationship). I am minded his purpose may have been to give independent scrutiny of the effectiveness of LetterWise but I have no knowlegde whatsoever of the extent of his independence or otherwise to Eatoni or LetterWise.Djm-leighpark (talk) 08:22, 19 May 2019 (UTC)


 * Delete Fails WP:GNG and fails seems like an advertisement. Syndicater (talk) 01:29, 13 May 2019 (UTC)
 * Note: This discussion has been included in the list of Companies-related deletion discussions. MarginalCost (talk) 04:23, 13 May 2019 (UTC)


 * Comment: * Delete: Not a sporting event, pop star, video game etc. Put it out of its misry quickly. Djm-leighpark (talk) 10:19, 13 May 2019 (UTC) I ask myself is this an unethical comment?
 * Weak Delete - Does not satisfy software notability. Contains promotional language.  Robert McClenon (talk) 17:11, 13 May 2019 (UTC)
 * To quote Dorcas Lane ... My one weakness is ... that I naturally ignore sales pitch language. Well at least I think I do.  If the wordwise entry makes things worse feel free to say so and back it out.  The article has received no love whatsoever.  Reading about Letterwise/Wordwise there was the significant MacKenzie/Kober/Smith3/Jones/Skepner paper "LetterWise: Prefix-based Disambiguation for  Mobile Text Input" which is quite often cited by other research.  Efforts to market it for Mobile(cell) phones seem to have failed with T9/Mulit-tap etc retaining the market referred to in More power to the thumb Economist 22 june 2002 p11 & 9783039114511 pp86.87. However there was some traction in DECT/cordless and Iridium satellite phones probably due to memory.  Eatoni got into a massive lawsuit with RIM(Blackberry) following another one and a make up before that.  More recently Eatoni seems to have continued with some phone apps and multiple language stuff and  is interesting but not sure how much is egged.  Most of the isn't in the article.   I might try tweaking the toning the apparent usage a bit ... if I make things worse revert it.Djm-leighpark (talk) 20:57, 13 May 2019 (UTC)  ✅ ... knowing me not perfect but have highlighted where claims were being made.Djm-leighpark (talk) 21:52, 13 May 2019 (UTC)  Now ❌ as assessed as spam below. Djm-leighpark (talk) 23:08, 18 May 2019 (UTC)


 * delete — not sure if this is notable or not, but the article as written is just spam. Vectro (talk) 20:51, 18 May 2019 (UTC)
 * @ ... I have had no option but to revert my contributions to the LetterWise article as they are obviously not helping and it is inappropriate for me to leave them in place with your assessment. Thankyou.Djm-leighpark (talk) 22:20, 18 May 2019 (UTC)
 * *I've a review. I can't be sure just spam is incorrect.  In fact it is probably an incorrect statement.  I think I had identified and moderated over-egged claims.  While I can see you are not WP:NOTHERE, you does useful contributions to e.g. WP:TEA and WP:NOTNOTHERE allows peoples to go round simply nominating and contributing to WP:APD I'm reasonably sure a good mix between the two isn't particularly healthly.  I note you are unclear about the notability and also aware WP:AFDISNOTCLEANUP so perhaps a despamming pass is necessary at some point.Djm-leighpark (talk) 13:39, 19 May 2019 (UTC)

 Relisted to generate a more thorough discussion and clearer consensus.
 * Keep: Source Scott MacKenzie is not critical but Frehner, Wigdor/Balakrishnan, Isokoski, Alavi have merit (I think one or two of those are passing only from memory but have independence); from the Eatomi/RIM suings of the 2000s we have the references under robert above, as also we have the reference for the N'Ko script language use. So we have a fair bit around one way and another.  The MacKenzie paper is a Goto with a lot of citations but the majority of those will be passing mentions.Djm-leighpark (talk) 13:39, 19 May 2019 (UTC)  Actually RC will be relatively good judge if I've eased the promotional language.  Where I added Multi-tap to the example it is I feel quite a nice example however the word sirs shows Letterwise in a more favourable light compared to how mama would show Multitap to its best advantage. Djm-leighpark (talk) 13:48, 19 May 2019 (UTC)

Please add new comments below this notice. Thanks, So said The Great Wiki Lord. (talk) 17:23, 19 May 2019 (UTC)
 * Comment: The LetterWise article has undergone further improvement over the past few days with additional references added some of which I believe sate WP:RS and pragmatically I would now find it hard to believe any lack of notability claim against the article as it now stands would succeed. I proffer that most promotional claims made in the article have been eliminated, qualified or moderated however my deepening scrutiny suggest some may persist (though they are now not serious in the overall context of the article).  For that reason particularly I am using an under construction while scrutiny, refinement and improvement of the article continues as a background task.  Thankyou.Djm-leighpark (talk) 10:56, 22 May 2019 (UTC)

Keep per the significant coverage in multiple independent reliable sources.    </li> <li></li> </ol>

<ol> <li> The article notes: "Eatoni Ergonomics, a small company based in New York, has developed a rival predictive-text technology called LetterWise which, it claims, is far superior to T9. Unlike T9, it is not based on a dictionary, but on a series of tiny rules that enable it to predict, given what has already been typed, which letter the user is most likely to want next. For example, “Q” is most often followed by “U”. Of the three letters “MNO”, words most commonly start with “O”. The letters “ST” are most often followed by a vowel. And so on. Like T9, LetterWise has a special key, used to step through possible letters when it guesses wrong. It may not sound like a great leap forward, but it works well. Unlike T9, LetterWise can be used to type proper names, addresses and Internet locations. It is simple to learn, and speeds of 38 words per minute (wpm) are possible, compared with 23 wpm for T9 and 20 wpm for multitap, says Howard Gutowitz, Eatoni's boss. LetterWise has benefits for phone makers as well as users. Being based on rules, not a dictionary, it occupies only three kilobytes of memory; T9 needs as much as 100 kilobytes. Many phones contain dictionaries for up to 20 languages, so this saving can be multiplied several times over. For Asian languages, the savings are even greater. And since mobile phones typically have only 2,048 kilobytes of memory, this leaves more room for games and other features. LetterWise's small size also makes it feasible for use in short-range cordless phones, which typically have no more than 128 kilobytes of memory. Yet despite all these advantages, LetterWise has failed to take off. Mobile-phone makers are reluctant to break ranks and defect from T9, says Mr Gutowitz. But he has a plan. LetterWise has been licensed to Philips, Siemens and Panasonic who, together, account for more than half the European market in cordless phones. Over the next few months, it will become possible to send text messages from fixed-line phones as well as mobile ones, and those phones will have LetterWise built in. Mr Gutowitz hopes this will enable the technology to reach a wide audience, and encourage mobile-phone makers to include it in their handsets, too. A deal with one mobile firm, he says, is already on the cards."</li> <li> The article notes: "How does it work? Both products use a standard mobile phone layout keyboard, so needn't change form factor at all, although the keys will need to be appropriately labelled. Letterwise is a predictive text entry system, and Wordwise is a more advanced product more likely to appeal to touch typists, and capable of greater speed. You can get a better handle on how they work if you understand what they're not. So, although Letterwise feels similar to the multitap systems common on mobile phones, where you press the key repeatedly in order to get the correct letter, Letterwise takes a stab at guessing the letter that should come next. So it'll kick off a word with the most common letter obtained via the key you press, and get more accurate as the word continues. If it gets the letter wrong (most commonly at the start of the word), then you countermand it by pressing the designated 'Eatoni' key, which The Register feels should be labelled the Nope key, but which currently isn't. What's happening here is that Eatoni is using predictive smarts, whereas multitap is simply using brute force. Your brute force, phone manufacturer's cost saving. Dictionary-based systems such as T9 are the more upmarket rivals, but Eatoni claims a 15 fold advantage in terms of queries and lookup errors over T9, and more crucially, dictionary-based systems take up an awful lot more space, meaning more cost for the phone manufacturers, and less likelihood of them becoming ubiquitous. Letterwise, on the other hand, uses a scalable database that can fit into a few kilobytes, although accuracy is better, the more storage there is. The database itself is not a dictionary, but a language-specific prediction system, and the most common letters used in English on the phone keypad are cehlnsty, since you ask. Frequencies differ from language to language, and Eatoni offers numerous different versions, including Latin. Why Latin? Because we could, apparently."</li> <li> The book notes: "Prefix-Based Disambiguation. To avoid the problem just noted, Eatoni Ergonomics (www.eatoni.com) developed an alternative to dictionary-based disambiguation. Their method, called LetterWise, uses prefix-based disambiguation [7]. Instead of using a stored dictionary to guess the intended word, LetterWise uses probabilities of “prefixes” in the target language to guess the intended letter. A prefix is simply the letters preceding the current keystroke. Implementations currently use a prefix size of three. For example, if the user presses the 3 key with prefix ‘_th’, the intended next letter is quite likely ‘e’ because ‘_the’ in English is far more probable than ‘_thd’ or ‘_thf’. The distinguishing feature is that prefix-based disambiguation does not use a dictionary of stored words: it is based on the probabilities of letter sequences in a language. Thus, the technique degrades gracefully when confronted with unusual character sequences, as in abbreviations, slang, etc. Switching to an alternate entry mode is not needed. Still, the wrong letter is occasionally produced, and in these cases the user presses the NEXT key to choose the next mostly likely letter for the given key and context. ... Although the comparison with dictionary-based disambiguation is less impressive, the KSPC figure for LetterWise does not carry the same assumption with respect to dictionary words."</li> <li> The book notes: "The LetterWise14 system addressed the T9 issues by using character N-grams rather than a dictionary. When the user presses a key, the most probable character, given the previous N-1 characters, is displayed. This means that the user sees immediately the character that has been selected. If this character is incorrect, Multitap techniques are used to get at the correct character. This is an extension of the earlier LessTap15 system that uses only character probabilities. LessTap and LetterWise are not reliant upon words being in the dictionary. MacKenzie et al have reported a KSPC of 1.15 for LetterWise. They also performed actual text entry experiments over many days to address the learning issues. In the first 25-minute session, LetterWise achieved 7.3 words per minute with Multitap generating 7.2 wpm. By the 20th session, LetterWise users achieved an average of 21 wpm with Multitap achieving 15.5 wpm."</li> <li> The article notes: "LetterWise is the first generation of Eatoni's LOPA software. It is designed to compete with the multi-tap systems in common use on mobile phones today. On average, multi-tap requires 2.2 keystrokes per letter typed, Gutowitz said. LetterWise, which eliminates the need for users to press a 'next' key or wait for a time-out, averages 1.18 strokes per letter. LetterWise uses three kilobytes of memory, small enough to fit easily onto a Subscriber Identity Module card. Although SIM cards generally are associated with GSM air interfaces, the predictive text system is agnostic regarding RF technology, Gutowitz said."</li> </ol>

There is sufficient coverage in reliable sources to allow LetterWise to pass Notability, which requires "significant coverage in reliable sources that are independent of the subject". Cunard (talk) 08:36, 26 May 2019 (UTC)</li></ul>
 * I don't think all of those pass WP:GNG - for instance, the MacKenzie scholarly article is fundamentally self-published - but I admittedly didn't see the Economist article in my before search. SportingFlyer  T · C  21:46, 29 May 2019 (UTC)

<div class="xfd_relist" style="border-top: 1px solid #AAA; border-bottom: 1px solid #AAA; padding: 0px 25px;"> Relisted to generate a more thorough discussion and clearer consensus.

Please add new comments below this notice. Thanks,  Sandstein   09:30, 26 May 2019 (UTC)
 * Keep Since this article is under major construction work.Catfurball (talk) 20:16, 29 May 2019 (UTC)
 * Comment:(As construction template steward) To a degree most if not all or nearly all sources relevant to notability are in place from an AfD viewpoint. The main reason its up there is I'm still tweaking and cross checking if a significant claim is made it is from a reliable source; this particularly affects the WordWise section.  I'd like to have had the template off by now or at least down to two sections.  Two mornings in a row I got up with the intention of concentrating on this but it hasn't happened and three mornings ago working back from another paper I needed to introduce and look at the (Primary) Kober paper.  The article is probably now at the point no-one would attempt to take it to AfD and most would not contemplate it.  However a sustained argument by a skilled possibly non neutral deletionist could make a defence by an unskilled person as myself difficult (witness TOPCAT (software) and Kst (software)).  Currerntly I feel have countered the initial argument to what I believe is this required level and that is unchallenged of recent and the article has now also has its subject broadened giving a wider notability surface.  Thankyou. 21:39, 29 May 2019 (UTC)  I observe the nom. has just altered content in the previous section as soon as I have made this comment selectively dismissing some of 's suggested source while sweeping over others.  I find it hard to believe MacKenzie's conference paper, experiments and study, widely cited, and at least partially verified was not scrutinised prior to conference acceptance.  But I have already noted 10 days ago MacKenzie et al is not critical for notability.  I note particularly Olsen is source not in the article that is useful from all points.Djm-leighpark (talk) 23:10, 29 May 2019 (UTC)


 * Keep. Per and WP:HEY.  The article is now suitably referenced (as noted above; and on reading it), and still doesn't yet include the citations that a search of books provides (one of which I added just now). A major improvement from where it was at the nomination, and clearly an interesting product that should be WP:PRESERVED in WP. Britishfinance (talk) 08:34, 2 June 2019 (UTC)


 * The above discussion is preserved as an archive of the debate. <b style="color:red">Please do not modify it.</b> Subsequent comments should be made on the appropriate discussion page (such as the article's talk page or in a deletion review). No further edits should be made to this page.