Slow as a Snail
“by persistence the snail reached the ark”
That’s the pace am in. Jeez, typing these transcription is taking too long than I expected. Well, am talking about transcribing the Malayalam word to the previously mentioned Phonetic representation. See for yourself.
അവസ്ഥ A V S3T2H
അവിചാരിത A VI CA
അവിടം A VI T1M
അവിടുത്തെ A VI T1U TT2AE1
അവിടുന്ന് A VI T1U NN2
അവിടെത്തന്നെയാണ് A VI T1AE1 TT2 NN2AE1 YA N1U1
അവിടെത്തെ A VI T1AE1 TT2AE1
അവിടെന്ന് A VI T1AE1 NN2
അവിടെയല്ലാ A VI T1AE1 Y LL1AA
അവിടെയാണ് A VI T1AE1 YA N1U1
അവിടെയായിരിക്കും A VI T1AE1 YA YI R2I KKUM
അവിടെയില്ലാത്തത് A VI T1AE1 YI LL1AA TT2 T22
അവിടെയുണ്ടെന്ന് A VI T1AE1 YU N1T1AE1 NN2
അവിടെയുണ്ട് A VI T1AE1 YU N1T1
അവിടെയുമല്ലാ A VI T1AE1 YU M LL1AA
As I transcribe the One Thousand Two Hundred and Forty Third word, I remember I haven’t shared the week with you guys. Well to tell the truth, this week was pretty slow and it's been raining. You know what’s coming. (** sighs **)
Yup, guessed it right. Again, erratic power failure (** aargghhh **).
After couple of days!
Well the reason I am writing this in a slow pace is because the week or the past 2 weeks have been very slow but its gaining its pace. Slow in the sense that the work am doing is kind of repetitive.
Each day, I sit in front of my PC, open up the Phonetic dictionary file, and start writing more phoneme transcriptions.
Blogging this over and over is going to be boring, but some new things are shaping up.
For GSOC to be successful like every other guys on forums and chats have said to me, the main thing to focus is on “how not to get distracted from coding”.
With the kind of work am doing, this is pretty simple. I mean to get distracted is pretty simple. ( ** irony duh ** ). Especially when there is rain happening outside your window ( ** aww…the feel of rain ** ) and you can’t do anything except muster any abuse at the state of power supply during the rainy time.
June 10 — update continues
Jokes aside, that’s how the weeks have passed away after the last update. But, as there is a mid term evaluation just around the corner, I know that I have to be focused (** I am! **), plus the organization and my mentor, have been kind enough to notify the same through e-mail (** kudos **).
And so, the fire picks up the rate and spreads. (** duh **), well it’s time to take some contributions towards collecting audio data for training. Now there are several ways to collect data from the Internet but to have a control over what we are developing, the simplest and feasible way is to record whatever we need in the correct format. In this way the recorded sounds can be used in the proper format that the tool requires it to be in. Problem with downloading pre recorded data from the Internet is that the audio format might be different from what we actually need. Even if the format is correct there is no guarantee that it might work for us.
In light of this conclusion I decided to take contributions from friends and the community so that the project itself has the meaning of being open source. (** yaay **)
Since the number of sentences in the collected source which I mentioned in my previous blog is about 5000 sentences, having around 20 different speakers each of whom will record about 250 will produce ample amount of speech data for the tool to be trained. Some of my friends ( Akhil, Sijin, Jithin) have already started contributing towards this through GitHub profile.
If you wish to contribute, head over to the repository and fork it. Then ping me. (** wink **)
puts "until then ciao!!"