One more voice added to the mix :-)... Now the bot supports TTS to German as well!!! Same commands, different language:

• start a English to German conference: say 'hello' and then select 'de' to activate German
• at any point type bot*say and then the sentence you want to translate and listen - for example 'bot*say And you? Do you speak German?'
• wait until the bot replies back with a pointer to a wave file and click the link.

I already bought the French voice as well, but the code share some portions with the Greetings parser, so I'll have to add the French localization files in order to use TTS. I'm considering MT'ing the sentences for now, until some real French speaker can help me with better sentences.

There you have it... Enjoy!

Adicionei suporte para text-to-speech em Alemao. E ao que parece, a voz e' bem legal tambem! O procedimento e' identico ao que eu implementei para Espanhol.

• comece uma conferencia de Ingles para Alemao: diga 'hello' e depois selecione 'de' para ativar Alemao
• a qualquer hora durante a conversa, digite bot*say seguido da frase que voce quer traduzir e ouvir - por exemplo 'bot*say And you? Do you speak German?'
• espere ate' que o bot responda com um ponteiro para um arquivo wave  click no link.

Eu tambem comprei a voz para frances, mas como o codigo compartinha alguns modulos com o modulo de traducao eu preciso implementar o Greeting para habilitar Frances antes de qualquer outra coisa.

Como sempre, divertam-se!

I guess you guys all know by now that Microsoft reveled the new IE8 during the Mix '08 conference, and also released the Beta bits for download (the XP version can be found @ http://www.microsoft.com/downloads/details.aspx?FamilyId=1A2E3DDD-B38B-439D-BBA7-F179A5D3ECAF&displaylang=en). I must confess I haven't had time to fully explore what's new, but I can talk about one REALLY cool feature called Activities. Activities are metadata driven extensions open to developers. Alright, I'm starting to sound like a Product Manager, so let's try this again: think about a way you could tell IE what to do when you highlight some text, hover over some link of simply right-click on a page. Now assume that instead of writing code to do that, you could instead put the instructions together in a XML file that would then be used to re-configure IE. Also, assume that you could then deploy this XML to any computer you want, and use a little Javascript snippet to allow users to import it themselves. That's how flexible Activities are...

Our team was selected to ship one of the standard activities 9the ones ypu get out of the box with IE 8) - a translation activity, of course. With IE 8 you can now translate selected text, translate the whole page or even point to a link and translate the corresponding target! It couldn't be easier, really... Just right-click and select 'Translate with Windows Live' from the menu.

Here's a quick summary of each feature:

• Translate Selected Text - Preview: you can preview the translation by selecting the text and hovering over the menu option (without clicking it). After a while a small window will appear, close to the menu item:

You can chance the language directly from the Preview window clicking 'Change Language' (duh!).

• Translate Selected Text - Execute: If you click on the menu item instead of hovering over, you'll be taken to the Translator main page, and all fields will be properly populated:

Note that the Url text box is pre-populated with the page you were browsing before, so if you decide to translate the whole page all you have to do is click the button ;-).

• Translate the whole page - Execute: without selecting any text, right-click anywhere in the page and click on the menu item. You will invoke the BilingualViewer interface, and the page will start to get translated
• Translate Links - Execute: select a hyperlink, right click to invoke the menu and click on 'Translate with Windows Live'. You will invoke the BilingualViewer interface, and the hyperlink target page will start to get translated
• Translate the whole page/Translate Links - Preview: the preview feature will only allow you to change the language selection. Hovering over the menu item you will see this:

This is pretty exciting news for us. It basically means that everyone with the new browser can have instant access to our translation services! We couldn't ask more in terms of getting more traffic, could we?

You can find more the new IE 8 features @ On10 - http://www.on10.net/blogs/larry/First-Look-Internet-Explorer-8/ 

Another cool feature brought you by "the little team that ships" - Microsoft Research Machine Translation (MSR-MT) !!!

Enjoy...

Bom, voces ja' devem saber que a Microsoft lancou a nova versao do Internet Explorer (IE 8) na conferecia Mix'08 a algumas semanas atras. Eu coloquei o link pra versao XP no blog acima, para quem quizer testar o Beta. Ainda nao tive tempo de entender todas as novidades, mas posso falar com seguranca a respeito de uma das varias features novas: as atividades de traducao. Com IE8 voce agora pode traduzir um texto selectionado, traduzir uma pagina inteira ou traduzir a pagina de um link com um simples click do mouse!

As novas atividades de traducao sao ativadas a partir do menu do right-click do mouse. Eu copiei varias telas no blog em Ingles que devem ser auto-explicativas - a unica opcao menos intuitiva e' a funcao 'Preview' que permite uma traducao super-rapida de um texto selecionado: basta selecionar o texto, usar o right-click e deixar o ponteiro do mouse sobre a opcao do menu (sem clicar). Depois de alguns segundo, uma janelinha ira' aparecer com o texto traduzido para o idioma que foi selecionado na ultima operacao do tradutor. Para trocar o idioma basta clicar em 'Change Language' (ou 'Alterar Idioma' :-) ).

As atividades sao hyper uteis e muito facil de usar. Acho que o trafego para nosso sistema vai aumentar horrores - basta pensar que agora todo usuario do IE8 tem a chance de disparar traducoes com um clique!

O time esta' de parabens - a feature ficou muito legal... E sai da frente que tem mais inovacao acontecendo em breve... E' so' aguardar!

Abracos

It's been a while since I got that excited about the project... Today I was able to finish the TTS implementation for multiple voices, and the bot can now speak with in both English and Spanish! And boy! The Spanish voice is really good...

Try this: invite the bot and start the conversation in English (say "hello"). Select "es" for Spanish and try something like bot*say And you? Do you speak Spanish? Give the bot some time, and you'll the the pointer to my Comcast personal site.

It works really really well!!! Too bad these voices are so expensive - I'd love to buy more voices for other languages but it's not a cheap hobby, you know...

Anyway, go practice your Spanish!

Modestia a parte, essa ficou legal... Consegui implementar o suporte para multiplos engines de TTS e instalei a voz para Espanhol que eu comprei ontem. Agora o bot pode falar Ingels e Espanhol, e posso garantir que a pronuncia da voz para Espanhol é fantastica!

Para ouvir a voz em Espanhol voce tera' que comecar uma conversa em Ingles e selecionar "es" . Tente algo facil para comecar: tipo bot*say And you? Do you speak Spanish? Em pouco tempo o bot responde com um arquivo Wave.

Se as vozes nao fossem tao caras eu bem que comprava mais... infelizmente, nao é um brinquedinho barato nao... Vou adicionando mais quando puder bancar, prometo.

Abracos...

I did not realize I missed a couple of conversions when I migrated the Bot to this new environment... As a consequence you could have been seeing a lot of question marks in the middle of the sentences. It should be fixed now - my bad.

On the good news front I got a decent SAPI TTS voice to replace Microsoft Sam. I also got a Spanish female voice, so I hope I can get it up soon as well. it took me a while to find the correct way to set different voices using SAPI, so here's the code in case someone is looking for a similar solution:

(assuming voiceDesc is the name of the voice to be used)

SpeechStreamFileMode spFileMode = SpeechStreamFileMode.SSFMCreateForWrite;
SpFileStream spFileStream = new SpFileStream();
spFileStream.Open(filename, spFileMode, false);

SpeechVoiceSpeakFlags spFlags = SpeechVoiceSpeakFlags.SVSFlagsAsync;
SpVoice Voice = new SpVoice();

// select the right voice for this session
foreach (SpeechLib.SpObjectToken aVoice in Voice.GetVoices("", ""))
{
    if (aVoice.GetDescription(0).StartsWith(voiceDesc))
    {
        Voice.Voice = aVoice;
        break;
    }
}

Voice.AudioOutputStream = spFileStream;
Voice.Speak(local, spFlags);
Voice.WaitUntilDone(10000);
spFileStream.Close();

Eu nao percebi que quando migrei o sistema para esse novo servidor eu deixei de incluir algumas conversoes. O erro deve ter gerado um monte de traducoes com pontos de interrogacao no meio dos textos - foi mal.

Pra compensar, comprei duas vozes TTS - uma para Ingles (US) e outra para Espanhol. Ja' instalei a voz para Ingles, que e' MIL vezes melhor que o Microsoft Sam. Eu tive dificuldade em encontrar o jeito correto de chavear entre vozes, entao pra facilitar a vida de quem esta' com o mesmo problema segue um snippet do codigo ai em cima.

I guess this is going to be a long post - I have tons to say. First things first: the bot is up and running, and it now has a new host. I've installed it on my new home server. It's a pretty darn good machine - Quad proc with 8Gbytes, running Win2003 64 bits.

 
New MTBot host

So what changed, besides the host? A couple of things:

• There was still some confusion about the project ownership, and the fact that the folks in the team where kind enough to add a demo during TechFest somehow added to it. The bot is still my own personal project - nothing changed on this front. The demos was meant to show how our MT systems can integrate with other Microsoft applications, Messenger been one of them.
• Some people where concerned to see IM messages passing thru Microsoft network hardware (remember, the bot was running on a machine under my office desk at work). Well, I'm not going to question that. Since the bot is my project, it better be running from outside Microsoft.
• There was also some concern about the fact that I could be using backdoors and other Microsoft internal tools/components in the project. Yes, I was accessing a test server that is not accessible from the outside, and yes, I was using a Text-To-Speech voice that is not a commercial product. I'm now using the MT production environment (http://translator.live.com)  so every sentence sent to the bot is translated using automation that emulates someone typing a sentence and click the 'Translate' button. I haven't found a decent alternative to the TTS voice yet, so you'll hear the charming 'Microsoft Sam' for now.
• No one seems to get the ':' separator in the bot special commands !!! I've added a '*' as separator as well so the bot now takes bot:bye, bot*bye, bot:say, bot*say.

Bringing the bot home - literally - addressed all issues, but you need to understand the following:

• my home network is fast, but it's no T1...
• if Comcast goes down, the bot goes down
• if power goes down during any storm, the bot goes down

Anyway, let's see how this new phase of the project goes - it's getting interesting by the day :-)

Eu instalei o bot em um novo servidor em casa - uma maquina bem legal com um Quad proc e 8Gbytes, rodando Win2003 64 bits. Tire a foto hoje.

A mudanca foi meio de repente... A motivacao principal foi a confusao que a demo na TechFest causou - tinha gente achando que o bot era oficial. Para evitar que a coisa se tornasse um problema, resolvi que era melhor isolar por completo qualquer interface que o bot pudesse ter com a Microsoft. Dessa forma nao tem como ninguem ficar paranoico porque as mensagens estao passando por routers e cabos dentro da grande empresa... E infelizmente, tive que abandonar a voz do TTS que eu estava usando (nao era um produto comercial). Por enquanto voces so' ouviram o taquara rachado que é oo Microsoft Sam.

Tive que alterar algumas coisas no projeto para conseguir rodar sem utilizar um ambiente de testes como retaguarda: agora toda traducao que o bot executa emula um acesso a pagina principal do MT (http://translator.live.com) e um clique no botao 'Traduzir'. Outro detalhe foi que eu percebi que ninguem presta atencao ao ':' que separa os comandos especiais do bot. Adicionei '*' como alternativa, portando o bot agora aceita bot:bye, bot*bye, bot:say, bot*say.

Bom, lembrem-se de duas coisas:

• a minha conexao em casa é rapida, mas nao é nenhuma T1
• se a Comcast cair, o bot cai...
• faltou luz, o bot cai...

Por enquanto as coisas estao indo bem - vamos ver como essa nova fase do projeto vai evoluir...

I've been really busy the last couple of days preparing for (guess what?) TechFest 2008! The bot made it to the demos! We're on booth 302 -"Bilingual Built-ins That Break Language Barriers".

It's been a while since I ran demos - actually about 15 years! - and it was cool to see all the attention and the noise our stuff generated. Aside from the the bot, the demos included:

• A very cool News Portal that allows users to search articles in many languages in parallel, providing translations for all headlines and linking them to our Bilingual Viewer UI.
• A prototype that uses our MT service to translate video captions
• The new IE 8 Translation Activities - I'll definitely add a new blog entry dedicated to it, hopefully with a video
• Our main UI - http://translator.live.com - and all other hooks to the translation system

Most of the feedback we got was really positive, and for all I can tell people get that we're working of two distinct fronts at the same time - improving the translation quality while adding more hooks to other Microsoft products and services. On the bot front, I'm really excited about the overall reaction - of course, some expected it to be able to understand the LOLs and BRBs and translate slang properly and (needless to say) were a bit disappointed with the experience :-).

Several people I know stopped by and paid us a visit, including Lorenzo (http://lorenzomadrid.spaces.live.com/).  In summary, it was a really busy day...

Eu estive tao ocupado com os preparativos do TechFest que até esqueci de anunciar que o bot foi incluido nas demos do grupo. Estamos no booth 302-"Bilingual Built-ins That Break Language Barriers". Muito bom! To hyper orgulhoso...

E so hoje me toquei que ja' faz um tempao que eu participei da minha ultima demo em feiras - acho que a ultima foi a uns 15 anos atras! Tinha esquecido de como era confuso - o nosso booth esteve cheio a maior parte do tempo. Além do bot, as outras demos do grupo incluem:

• Um portal bem legal que permite que os usuarios executem buscas em varios idiomas em paralelo. Os resultados sao grupados por manchete e por idioma, e tudo e' conectado direto ao BilingualViewer
• Um prototipo que conecta legendas do Media Player com nosso sistema de traducao
• Uma nova feature do IE 8 chamada Transaltion Activity - vou bloggar sobre isso depois do TechFest
• e a claro, nosso produto principal - http://translator.live.com - e todas as interfaces disponiveis

A maioria do feedback foi positivo, inclusive sobre o bot.nosso time é muito pequeno, e fica dificil trabalhar em duas frentes ao mesmo tempo. Mas no geral, estamos indo bem...

Muita gente apareceu por la', inclusive o Lorenzo. Foi um dia bem corrido, mas valeu...