Repository navigation
Handle Full Text Search phrases outside of individual segments. #179
Description
Activity
- added a commit that references this issue
on Sep 13, 2024 Proposed solution to finding time stamps from snippet referenced in #178
def find_phrase_indexes(phrase, arr): marks = [] fullText = [] for i, row in enumerate(arr): for word in row[2].strip().split(): marks.append(i) fullText.append(word) ans = [] curr = 0 phraseArr = phrase.split() for i, search in enumerate(fullText): if search == phraseArr[curr]: curr += 1 if curr == len(phraseArr): ans.append([marks[i-len(phraseArr)+1], marks[i]]) curr = 0 return ans
The method of returning keywords doesn't work well when the users searches for a phrase that exists across separate rows within the transcript. It also doesn't provide much context. For example if the query is
foo bar pancake yeetand the rows are['foo bar ', 'pancake', 'yeet']the current method will return three rows and offer little context. If we create a dedicated table for the full transcript and combine it with the fts5 snippet function we can allow the user to control the number of words around their search keyword and still provide precise time stamps. We still need a way to find the time stamps of the returned snippet given a segment of arbitrary length.