You are here
Sandhi Statistics
Presented below are the results of an automated sandhi analysis of this Chapter of the Srimad Bhagavad Gita. All the sandhi sutras applicable to this Chapter, as well as one example of each sutra, are listed on the next page.
As discussed in the Sandhi Statistics of Chapter 1, a grammatical analysis of a Sanskrit sentence or stanza must analyze every word in a sentence, as Paninian sandhi sutras transform (or choose not to transform) underlying terms (declensions, conjugations, and indeclinables) into 'single words' as well as 'combination words'. Hence, we use the broader term 'sandhi analysis' instead of 'sandhi splitting'.
| # | Description | INPUT | OUTPUT | ||||
| FALSE | FALSE | ||||||
| CORRECT | NEGATIVE | TOTAL | % | POSITIVE | |||
| A | Combination words (two or more terms) | 86 | 200 | 2 | 202 | 49.5 | 0 |
| B | Changed single-word terms | 93 | 92 | 1 | 93 | 22.8 | 2 |
| C | Unchanged Vocatives / Special terms | 38 | 37 | 1 | 38 | 9.3 | |
| D | Unchanged non-special single-word terms | 75 | 75 | 75 | 18.4 | 1 | |
| TOTAL | 292 | 404 | 4 | 408 | 100.0 | 3 | |
| Errors | <1% | 0.7% | |||||
The Columns labelled 'False Negative' and 'False Positive' are explained in the Sandhi Statistics of Chapter 1.
As will be noted from the above table, the number of errors in 'sandhi analysis' is reasonably small ( False Negatives <1%, False Positives <1%).
The following is a summary of the False Negatives and False Positives discussed in the table above.
| Stanza | False Negatives | False Positives |
| 7.2 | na, iha | neha |
| 7.14 | mAyA | mAyAs |
| 7.18 | sarve | sarvas |
A.
In Stanza 7.2, the sandhi analyzer incorrectly assumed that the combination word 'neha' was an unchanged underlying term 'neha' instead of the combination of the underlying terms 'na' and 'iha'. Please note that 'neha' is a perfectly valid underlying term (i.e. an exceptional conjugation of the verb 'nah' -- 'to bind' -- in the 2nd Person Plural 'perfect' conjugation), and the same error was also seen previously in Stanza 2.4. Clearly, these errors would go away if the 'sandhi analyzer' were to disallow the verb conjugation 'neha' -- but this decision would result in an error in case the correct answer is, in fact, 'neha' (and not 'na iha'). At a future stage, we may evaluate the trade-offs between rare terms and common terms. However, it is likely that the choice of the verb 'neha' may be detected in the next stage of parsing (i.e. the syntactic analysis), as an unexpected verb that causes a parsing failure. This error accounts for 2 False Negatives in Row A and 1 False Positive in Row D (i.e. 'neha').
B.
In Stanza 7.18, we come across another familiar error (seen earlier in Stanza 1.6 and Stanza 1.11) that is due to inherent ambiguity. The 'sandhi analyzer' incorrectly assumed that 'sarva' was derived from the underlying term 'sarvas' by the application of the 8.3.17-bho bhago agho apUrvasya yo'shi sutra (with the 8.3.19-lopaHa shAkalyasya elision sutra). However, it was actually derived from the underlying term 'sarve' by the application of the 6.1.78-echo'yavAyAvaHa sutra (with the 8.3.19-lopaHa shAkalyasya elision sutra). This error accounts for 1 False Negative in Row B and 1 False Positive in Row B.
C.
In Stanza 7.14, we come across another familiar error (seen in several of the preceding chapters) that is due to inherent ambiguity. The 'sandhi analyzer' incorrectly assumed that 'mAyA' was derived from the underlying term 'mAyAs' by the application of the 8.3.17-bho bhago agho apUrvasya yo'shi sutra (with the 8.3.22-hali sarveSHAm elision sutra). However, it was actually the unchanged underlying term 'mAyA'. This error accounts for 1 False Negative in Row C and 1 False Positive in Row B.
D.
This False Positive in Stanza 7.2 has been discussed under point A above.
We now see a familiar pattern emerging from the analyses in all the preceding chapters, of a few sandhi sutras being responsible for most of the ambiguity. Most of this ambiguity can be identified and rectified by a syntactic parser (that is run after the sandhi analysis) that is made aware that a certain underlying term is ambiguous. This process is not very different from the ambiguity resolution practiced by a human expert.