Sandhi software

[dhaval patel]

Dear all,

Please find attached link to a software for doing sandhi which I am currently trying to create. 

http://www.lanover.com/lan/sanskrit/sandhi.html

This is the place where it will be available for testing.

I have tried to include prakriyA as per siddhAntakaumudI text.

Wherever there is deviation, please feel free to drop a message either in this thread or offline. 

Please oblige me by spending some of your precious time and giving your valuable inputs.

N.B. The source code (PHP) used for this machine is available at https://github.com/drdhaval2785/sanskrit

-- 

Dr. Dhaval Patel, I.A.S

District Development Officer, Rajkot

www.sanskritworld.in

[Mārcis Gasūns]

Perfect!

It seems to be the first ever Sanskrit open source sandhi machine. After it will learn to generate sandhis, it will split them one day as well, I hope.



You entered: उद् + जल्

By jhalAM jazo'nte (8.2.39), The padAnta is 'jhal' is replaced by 'jaz' :

N.B.: If the jhal is at the end of a pada, it is mandatory to change it. Otherwise it is not going to change. Ignore which is not applicable.

1 - उद्जल्

---

By stoH zcunA zcuH (8.4.40) :

1 - उज्जल्

---

By anaci ca (8.4.47):

Please note: Wherever there is dvitva, it is optionally negated by sarvatra zAkalyasya. (8.4.51)

1 - उज्जल्

2 - उज्ज्जल्

---

By anaci ca (according to mahAbhASya example of vAkk) :

Please note: Wherever there is dvitva, it is optionally negated by sarvatra zAkalyasya. (8.4.51)

1 - उज्जल्

2 - उज्ज्जल्

3 - उज्जल्ल्

4 - उज्ज्जल्ल्

---

N.B.: By triprabhRtiSu zAkaTAyanasya (8.4.50), the dvitva is optionally not done in cases where there are more than three hals appearing consecutively. e.g. indra - inndra.

By halo yamAM yami lopaH (8.4.64) :

1 - उज्जल्

2 - उज्ज्जल्

3 - उज्जल्ल्

4 - उज्ज्जल्ल्

---

By jharo jhari savarNe (8.4.65) :

1 - उज्जल्

2 - उज्ज्जल्

3 - उज्जल्ल्

4 - उज्ज्जल्ल्

---

Final forms are :

1 - उज्जल्

2 - उज्ज्जल्

3 - उज्जल्ल्

4 - उज्ज्जल्ल्

---

[विश्वासो वासुकिजः (Vishvas Vasuki)]

On Tue, Nov 12, 2013 at 12:00 PM, Mārcis Gasūns <gas...@gmail.com> wrote:

It seems to be the first ever Sanskrit open source sandhi machine. After it will learn to generate sandhis, it will split them one day as well, I hope.

+1. I wish other grammar tools like tdil-dc.in/san/sandhi/sandhi_dit.html were open-sourced and published (eg: on github.com )

--
--
Vishvas /विश्वासः

[dhaval patel]

In Marcis's input,

halo yamAM yami lopaH was wrongly applied, and jhalAM jaz jhazi was missed out.

It was pointed out by Mr. H N Bhat.

It has been corrected and uploaded on web now.

You entered: उत् + जल्

By jhalAM jazo'nte (8.2.39), The padAnta is 'jhal' is replaced by 'jaz' :

N.B.: If the jhal is at the end of a pada, it is mandatory to change it. Otherwise it is not going to change. Ignore which is not applicable.

1 - उद्जल्

---

By stoH zcunA zcuH (8.4.40) :

1 - उज्जल्

---

By anaci ca (8.4.47):

Please note: Wherever there is dvitva, it is optionally negated by sarvatra zAkalyasya. (8.4.51)

1 - उज्जल्

2 - उज्ज्जल्

---

By anaci ca (according to mahAbhASya example of vAkk) :

Please note: Wherever there is dvitva, it is optionally negated by sarvatra zAkalyasya. (8.4.51)

1 - उज्जल्

2 - उज्ज्जल्

3 - उज्जल्ल्

4 - उज्ज्जल्ल्

---

N.B.: By triprabhRtiSu zAkaTAyanasya (8.4.50), the dvitva is optionally not done in cases where there are more than three hals appearing consecutively. e.g. indra - inndra.

By jhalAM jaz jhaSi (8.4.53):

1 - उज्जल्

2 - उज्ज्जल्

3 - उज्जल्ल्

4 - उज्ज्जल्ल्

---

By jharo jhari savarNe (8.4.65) :

1 - उज्जल्

2 - उज्ज्जल्

3 - उज्जल्ल्

4 - उज्ज्जल्ल्

---

Final forms are :

1 - उज्जल्

2 - उज्ज्जल्

3 - उज्जल्ल्

4 - उज्ज्जल्ल्

--
निराशीर्निर्ममो भूत्वा युध्यस्व विगतज्वरः।। (भ.गी.)
---
You received this message because you are subscribed to the Google Groups "भारतीयविद्वत्परिषत्" group.
To unsubscribe from this group and stop receiving emails from it, send an email to bvparishat+...@googlegroups.com.
To post to this group, send email to bvpar...@googlegroups.com.
Visit this group at http://groups.google.com/group/bvparishat.
For more options, visit https://groups.google.com/groups/opt_out.

[dhaval patel]

+1. I wish other grammar tools like tdil-dc.in/san/sandhi/sandhi_dit.html were open-sourced and published (eg: on github.com )

This is precisely the reason why everybody ends up reinventing the wheel. I see at least three sandhi machines already developed by JNU / HCU / TDIL. There may be many others. But as long as the information is in closed domain, there is no scope to modify it for one's individual need.

[विश्वासो वासुकेयः]

On Tuesday, November 12, 2013 9:39:41 PM UTC-8, amba kulkarni wrote:

 Dear Dhaval Patel ji and Vishvas ji,

The sandhi joiner developed by University of Hyderabad in collaboration with RS Vidyapeetha Tirupati is available under GPL.

नमस्ते अम्बार्ये। 

That is great news. Where is it published?
In general, the best practice is to provide a link on the tool's interface to the source-code behind it (as was done by shreevatsa here - http://sanskritmetres.appspot.com/ ).

If it is not published on the internet, I strongly recommend using the github.com repository popular among developers - It enables parallel development, and makes branching easy -- in addition the original authors can continue to use it for version control.

[विश्वासो वासुकिजः (Vishvas Vasuki)]

-others, + shreevatsa and sanskrit-programmers

On Tue, Nov 12, 2013 at 10:19 PM, विश्वासो वासुकेयः <vishvas...@gmail.com> wrote:

In general, the best practice is to provide a link on the tool's interface to the source-code behind it (as was done by shreevatsa here - http://sanskritmetres.appspot.com/ ).

Oops - sorry shreevatsa - forgot you asked us not to announce widely, but I forgot that in a separate discussion.

[dhaval patel]

On Wed, Nov 13, 2013 at 11:09 AM, Amba Kulkarni <ambap...@gmail.com> wrote:

 Dear Dhaval Patel ji and Vishvas ji,

The sandhi joiner developed by University of Hyderabad in collaboration with RS Vidyapeetha Tirupati is available under GPL.

Ma'am, I would be highly obliged if the source code can be shared. 

 

The same sandhi joiner is given to TDIL as a part of the consortium project, and is on TDIL website.

There is another sandhi joiner developed by Prof. Gérard Huet http://sanskrit.inria.fr

What is required is collaboration. In 2006 when I noticed that Prof Huet is also working in this direction, we started collaborating with each other to avoid any duplication of efforts. This also resulted in a consortium for Sanskrit Computational Linguistics, which has organised 5 conferences till now, and has led to many more collaborative efforts both at the national as well as international level. 

Prof Huet has already developed a very sophisticated practical sandhi splitter (segmenter),

I would love to see the source code / algorithm for the same. 

and at the university of Hyderabad, we have developed a parser (sentential analyser). Further, these tools also interact with each other, i.e. I can use heritage segmenter to split the Sanskrit sentence into words, which then can be passed on my parser for kaaraka level analysis.

I would love to see the source code / algorithm for both.

 

This is possible only through the collaboration.

I Agree. 

 

On the one side there is a necessity to join the hands together, so that there are no duplication of efforts.

On the other hand, the researchers may like to develop the same tool following different algorithms. For example, if we take English language, there are dozens of parsers for English being developed following different approaches.  Many of them are available either under open source, or for reasearch purpose, or as an online tool.

What I observe is many enthusiasts start the work on Sanskrit, but once they encounter the complexity the language poses, their enthusiasm dies.
 

Now coming back to Sri Dhaval Patel ji's efforts. I see that what he is trying to do is something different from the way I have done it.

He is trying to develop an algorithm, where he is ordering the sandhi rules manually.

The way I have implemented is just a simple ready-made matrix.  This I needed to invert the rules quickly.

Would you please elaborate what is meant  by 'simple ready-made matrix' ? 

 

But as a researcher I'll be more interested in knowing how the order in the Ashtadhyayi itself can trigger the order of the rules.

Would you please elaborate this ordering issue ? This is interesting.

 

Is it necessary to use 'paribhaashaa's such as 'nitya-antaranga...' etc.

Do you mean that the machine should be able to understand what is to be done when we tell it 'iko yaNaci' ? 

If yes, this is an interesting aspect. I have done a few functions in the code which are in a way telling the comp the same thing.

e.g. prat('hl') will tell the computer that it has to take the whole pratyAhAra starting from h and ending at l.

Similarly savarna (prat('Jl'),prat('cr')) will tell the computer to give the savarna of the pratyAhAra झल्‌ from the pratyAhAra चर्‌.

 

So I myself will be implementing sandhi joiner in a different way.

 

I would welcome more such efforts, which will revive the Paninian tradition,and at the same time I also welcome the collaboration, and sharing of resources to avoid duplication of efforts.

I would be sharing whatever I can do and keep the list posted about it.

The code is in nascent stage. There are around 50 known issues. I am trying to create a page where it will be proactively disclosed.

 

With regards,
Amba Kulkarni

On 13 November 2013 10:35, dhaval patel <drdhav...@gmail.com> wrote:

+1. I wish other grammar tools like tdil-dc.in/san/sandhi/sandhi_dit.html were open-sourced and published (eg: on github.com )

This is precisely the reason why everybody ends up reinventing the wheel. I see at least three sandhi machines already developed by JNU / HCU / TDIL. There may be many others. But as long as the information is in closed domain, there is no scope to modify it for one's individual need.

--
You received this message because you are subscribed to the Google Groups "samskrita" group.
To unsubscribe from this group and stop receiving emails from it, send an email to samskrita+...@googlegroups.com.
To post to this group, send email to sams...@googlegroups.com.
Visit this group at http://groups.google.com/group/samskrita.

For more options, visit https://groups.google.com/groups/opt_out.

--
आ नो भद्रा: क्रतवो यन्तु विश्वत: ll
Let noble thoughts come to us from every side.
- Rig Veda, I-89-i.
Assoc Prof.  and Head
Department of Sanskrit Studies
University of Hyderabad

Prof. C.R. Rao Road 

Hyderabad-500 046

(91) 040 23133802(off)

http://sanskrit.uohyd.ernet.in/scl
http://sanskrit.uohyd.ernet.in/faculty/amba

--
निराशीर्निर्ममो भूत्वा युध्यस्व विगतज्वरः।। (भ.गी.)
---
You received this message because you are subscribed to the Google Groups "भारतीयविद्वत्परिषत्" group.
To unsubscribe from this group and stop receiving emails from it, send an email to bvparishat+...@googlegroups.com.
To post to this group, send email to bvpar...@googlegroups.com.
Visit this group at http://groups.google.com/group/bvparishat.
For more options, visit https://groups.google.com/groups/opt_out.

[Mārcis Gasūns]

Namaste,

Dhaval's dream comes true. Code, that was not open for 10 years, is becoming open for the first time. 

A historical moment indeed. Copyrights say 2012, so no new changes in 2013, right? Very interesting.

M.

[dhaval patel]

Thank you very much ma'am.

It will take me some time to go through your code, because I don't understand the programming language used. But a plain reading of the code shows that it will not be difficult to understand that language.

Thank you indeed.

And to keep the discussion focused - 

The main idea is to find out the bugs in the system and repair it. 

So, any help in form of testing and reporting the errors will be highly appreciated.

I am attaching the known issues of the code as of now.

Any help to sort them out will also be highly welcome.

[विश्वासो वासुकिजः (Vishvas Vasuki)]

On Wed, Nov 13, 2013 at 6:27 AM, Amba Kulkarni <ambap...@gmail.com> wrote:

Dear Visvas,

  If somebody can put it in github.com, I'll appreciate the help.

नमस्करोम्य् अम्बार्ये।

I have uploaded uohyd code I received here:

https://github.com/vvasuki/sanskritnlp/tree/master/src/main/perl/uohyd

It includes the sandhi code you just sent, and the amarakosha data I got from you long time ago.

If you (or anyone else) have other code to share, please don't hesitate to let me know.

[विश्वासो वासुकिजः (Vishvas Vasuki)]

On Wed, Nov 13, 2013 at 4:12 AM, Dr. Ramanujan <ra...@cdac.in> wrote:

In fact, as part of our DESIKA software, Vedic accent sandhi-s and generation of Veda-vikrti-s were also part of C-DAC’s sandhi program in 1990.

Due to limitations of resources, change of platforms and lack of support/patronage they all cannot reach users.

नमस्करोमि रामानुजार्य।

That is really unfortunate.. If you still have the source code and associated documentation, however outdated, there is a small chance it will be useful to someone now or  in the future. Please publish it. If you like I can add it to a github repository. 

[Mārcis Gasūns]

On Wednesday, 13 November 2013 21:10:33 UTC+4, विश्वासो वासुकिजः wrote:

That is really unfortunate.. If you still have the source code and associated documentation, however outdated, there is a small chance it will be useful to someone now or  in the future.

Right, you never know.

 

Please publish it. If you like I can add it to a github repository. 

Yes, 1990 is not 1887, so the code could be just fine, to see the logic behind it. It's the algorithm that is valuable, not the code itself. 

[dhaval patel]

On Wed, Nov 13, 2013 at 10:42 PM, Amba Kulkarni <ambap...@gmail.com> wrote:

I looked at various kinds of issues you have noted down.

A similar study of suutras related to compound formation has been undertaken by my student.

He will be presenting his work at the ICON -- the conference on NLP at Noida in Dec 2013. After the conference, the paper will be available on my webpage.

There we classify the suutras into different categories and also discuss various kind of information needed by machine to generate the compounds automatically from the given paraphrase.

Will look forward to it.. Please let us know when that gets published.

 

-- Amba Kulkarni

Regards,
Amba

--

[dhaval patel]

Question 1:        Is there a consolidated state of the art review ( for the period 1970 to 2013)  on Samskrutham Language modeling for Computer related work – As an Indian Tradition ( beyond the Harvard school thoughts reflected in Sanskrit Computational linguistics analysis ?   To the best of my knowledge, the work in this area has been isolated and fragmented efforts, addressing only peripherals of Paninian rules. This is not sufficient to address any real issue related to Devanagari script or Samskrutham or the Indian languages needs.

I don't think there are many reviews of such sort. Before reviews, we should at least pull together the sources first. Once the sources are listed, the review can be undertaken by experts. A welcome step in this regard is taken by Vishvas by uploading many open source codes on his Github page. And the latest addition was by Prof. Amba Kulakarni. Once we have the resources, the review can go on.

 

 

Question 2:      How do we get  ‘collaboration’ with the Indian teams hear,  to explore < different approaches and algorithms> to pursue the issues  and deliver practical implementable  sustainable  solutions ?  Productivity and Research are not always best suited for  Open Source deliberation !

 

Once the review is over and there are specifically laid down requirements on what needs to be done, there are many sanskrit oriented programmers who can do this job on voluntary basis. I disagree with the presumption that productivity and research are not always best suited for open source deliberation. 

And by the way, sanskrit related softwares have always been developed not for commercial gain, but by personal interests. So, as far as sanskrit computational development is concerned, it can always be open source.

 

Hope the august members of the groups will do their level best to bring the hitherto efforts to open source domain.

Thanking you all.

[Mārcis Gasūns]

Namaste,

Everything about this software is great, even the GNU license. It is an era of open source for Sanskrit coming at least, it took it so long, no way back anymore.

I do not know much about sandhi. So I started comparing with my Sandhi table results https://www.dropbox.com/s/6iqo8w0qhkqtsa8/Sandhi-Table-19.10.13.pdf

If I open http://www.lanover.com/lan/sanskrit/sandhi.php directly it's empty. It better be not :)

I would want to link to some sandhi cases, but can only link to the sandhi page, not the cases.

3 - अहंंवलब्ह

As per me अहंंवलब्ह

has the same issue as Cologne dictionaries.

So a few days ago I wrote an email to them:

Please add chandrabindu to the converter.

MM is not 2 folowing anunasikas. 

MM is 1 chandrabindu.

ऊंं -> ऊँ

And I wonder how is

9 - अहंँव्वलब्ह

possible at all. So instead of

Final forms are :

1 - अहंवलब्ह

2 - अहंवलब्ब्ह

3 - अहंंवलब्ह

4 - अहंवलब्हँ

5 - अहंवलब्ब्हँ

6 - अहंंवलब्हँ

7 - अहँव्वलब्ह

8 - अहँव्वलब्ब्ह

9 - अहंँव्वलब्ह

10 - अहँव्वलब्हँ

11 - अहँव्वलब्ब्हँ

12 - अहंँव्वलब्हँ

13 - अहंवलब्भ

14 - अहंवलब्ब्भ

15 - अहंंवलब्भ

16 - अहंवलब्भँ

17 - अहंवलब्ब्भँ

18 - अहंंवलब्भँ

19 - अहँव्वलब्भ

20 - अहँव्वलब्ब्भ

21 - अहंँव्वलब्भ

22 - अहँव्वलब्भँ

23 - अहँव्वलब्ब्भँ

24 - अहंँव्वलब्भँ

my humble tables lists only 2.

It can be हंव or हँव्व. 2 instead of 24.

You entered: आन् + अज्

Final forms are :

1 - आनग्

2 - आनग्ग्

3 - आनक्

4 - आनक्क्

is as expected. So is

You entered: अन् + अज्

Final forms are :

1 - अन्नग्

2 - अन्न्नग् - ?

3 - अन्नग्ग्

4 - अन्न्नग्ग् - ?

5 - अन्नक्

6 - अन्न्नक्

7 - अन्नक्क्

8 - अन्न्नक्क्

Last try for tonight.

You entered: अक् + ह

Final forms are :

1 - अग्ह

2 - अग्ग्ह

3 - अग्हँ

4 - अग्ग्हँ

5 - अग्घ

6 - अग्ग्घ

7 - अग्घँ

8 - अग्ग्घँ

I would be happy with 2, but I get 8. Good or bad - not sure yet. Anyway thanks.

[nitish chandra]

Dhaval has mentioned some issues with this software. In 12th point, he has  mentioned शकन्ध्वादिषु पररूपं वाच्यम् is आकृतिगण, so naturally, some more forms can be included. But, just to test I queried पत + अञ्जलिः, using प्रातिपदिक+प्रत्यय option, and the only result was पताञ्जलिः using अकः सवर्णे दीर्घः . In सिद्धान्तकौमुदी, पतञ्जलिः is given as the result. Is this a bug?

Correct me if I am wrong.

[dhaval patel]

I took these words from bAlamanoramA as vigraha - so coded for patan + aJjali as the input for this zakandhvAdiSu pararUpaM vAcyam. This input should work and give proper output as pataJjali.

On Sat, Nov 16, 2013 at 11:30 AM, nitish chandra <nitishcha...@gmail.com> wrote:

Dhaval has mentioned some issues with this software. In 12th point, he has  mentioned शकन्ध्वादिषु पररूपं वाच्यम् is आकृतिगण, so naturally, some more forms can be included. But, just to test I queried पत + अञ्जलिः, using प्रातिपदिक+प्रत्यय option, and the only result was पताञ्जलिः using अकः सवर्णे दीर्घः . In सिद्धान्तकौमुदी, पतञ्जलिः is given as the result. Is this a bug?

Correct me if I am wrong.

--
You received this message because you are subscribed to the Google Groups "sanskrit-programmers" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sanskrit-program...@googlegroups.com.

For more options, visit https://groups.google.com/groups/opt_out.

[dhaval patel]

On Sat, Nov 16, 2013 at 4:15 AM, Mārcis Gasūns <gas...@gmail.com> wrote:

Namaste,

Everything about this software is great, even the GNU license. It is an era of open source for Sanskrit coming at least, it took it so long, no way back anymore.

I do not know much about sandhi. So I started comparing with my Sandhi table results https://www.dropbox.com/s/6iqo8w0qhkqtsa8/Sandhi-Table-19.10.13.pdf

If I open http://www.lanover.com/lan/sanskrit/sandhi.php directly it's empty. It better be not :)

Now, you may see a link "Go Back" on that page, which will take you to the HTML page.

 

I would want to link to some sandhi cases, but can only link to the sandhi page, not the cases.

Now, that has been fixed. e.g. http://www.lanover.com/lan/sanskrit/sandhi.php?first=iti&second=ucyate&tran=Devanagari&pada=pada . So now you can share the specific cases also.

3 - अहंंवलब्ह

As per me अहंंवलब्ह

has the same issue as Cologne dictionaries.

So a few days ago I wrote an email to them:

Please add chandrabindu to the converter.

MM is not 2 folowing anunasikas. 

MM is 1 chandrabindu.

ऊंं -> ऊँ

This depends on the input conventions. In sanskrit sandhis, there are cases where there are two consecutive anusvAras. So, this input convention may be dangerously wrong in that case.

The input doesn't look like sanskrit words.

That is based on the user preferences. The machine can duplicate. If there is popular demand, maybe there can be a user input which will ask the user whether they want to see the dvitva cases or not. Let me try for that.

 

--
You received this message because you are subscribed to the Google Groups "sanskrit-programmers" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sanskrit-program...@googlegroups.com.

For more options, visit https://groups.google.com/groups/opt_out.

[विश्वासो वासुकिजः (Vishvas Vasuki)]

On Sun, Nov 17, 2013 at 12:48 AM, dhaval patel <drdhav...@gmail.com> wrote:

In sanskrit sandhis, there are cases where there are two consecutive anusvAras.

कथमिव?

[dhaval patel]

विसर्जनीयस्य सः इत्यत्र व्याख्यानावसरे सिद्धान्तकौमुद्यामेवमुक्तम्‌ । अत्रानुस्वारस्यापि द्वित्वं भवति ।

--
You received this message because you are subscribed to the Google Groups "sanskrit-programmers" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sanskrit-program...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

[विश्वासो वासुकिजः (Vishvas Vasuki)]

कुतः स्वीकृतमिदं चित्रम्?

[dhaval patel]

Dear all,

Thank you for your continued support and feedback on sandhi machine based on siddhAntakaumudI prakriyA.

The revised website for testing after corrections is available at :

http://lanover.com/lan/sanskrit/sandhi.html

The revised source code is available at :

https://github.com/drdhaval2785/sanskrit

The testing bugs and known issues which were there in version 1.1 are documented at :

https://docs.google.com/document/d/1eWKKAwuWSQf0JbrFW21l9dcIue6N2CMlKtF5rT370qQ/edit

The known issues of current version 1.2 is documented at :

https://docs.google.com/document/d/1eWKKAwuWSQf0JbrFW21l9dcIue6N2CMlKtF5rT370qQ/edit#heading=h.j96h2xeb9tu6

Please feel free to send your feedback on list or off list.

(The editing to google docs is restricted because of security issues.)

If any member wants to contribute to the bug testing, the editorial rights to docs can be given happily. :)

[Mārcis Gasūns]

Namaste,

  Would it be possible to add batch function? To work with sandhis of a whole text, not just a few words?

M.G.