Solved

I have a string which has a conversation between two people along with their speaker tag.

I want to split the string into two sub strings containing speaker 1 and speaker 2 conversation only.

This is the code I am using to obtain the transcript.

operation = client.long_running_recognize(config, audio)
response = operation.result(timeout=10000)
result = response.results[-1]
words_info = result.alternatives[0].words
transcript = ''
tag=1
speaker=""
for word_info in words_info:
    if word_info.speaker_tag==tag:
        speaker=speaker+" "+word_info.word
    else:
        transcript += "speaker {}: {}".format(tag,speaker) + '\n'
        tag=word_info.speaker_tag
        speaker=""+word_info.word
transcript += "speaker {}: {}".format(tag,speaker)

This transcribes both speaker 1 and speaker 2 in the same file.

Solved: The Solution was much simpler. Thanks for the help.

transcript_1 = ''
transcript_2 = ''

for word_info in words_info:
    if word_info.speaker_tag==1:
        #speaker += " "+word_info.word
        transcript_1 += " " + word_info.word
    elif word_info.speaker_tag==2:
        #speaker += " "+word_info.word
        transcript_2 += " " + word_info.word

Depending of how do you get the data, I mean, if you get an unique raw string with all the messages from both speakers or you get the messages from each speaker separately.

A basic approach would be to establish the string "speaker X:" (where N is the speaker number) as the speaker tag for the first speaker then you could extract each message from each speaker using tools like NLTK and/or built-in functions like find().

Note: When I talk about a tag, I refer to some expression that would allow us to determine if a message is from a certain speaker or not.

Example: You get the whole text that includes all the interventions of the speakers.

• Steps to follow:

1) Set all speakers tags to distinguish their interventions in the whole text. Example: The speaker tag for the first speaker could be "speaker 1:"

2) Find all the interventions of a speaker using str.find("speaker_tag")

3) Add all the interventions of each speaker to different data structures. I think that a list of interventions of the speaker could be useful and then if you want to get all these interventions in one text message again, you could use some built-in function like str.join() to joining them into one string again.

Other option to solve this would be using a tool like NLTK (I think this tool is great to classify text)

It has very useful features like tokenization that I think it's would be useful to solve your problem.

In the following example, I am going to use find() and slicing for a basic example about text tokenization:

Text data:

text = "speaker 1: hello everyone, I am Thomas speaker 2: Hello friends, I am John speaker 1: How are you? I am great being here speaker 2: It's the same for me"

Code example:

from itertools import islice, tee

FIRST_SPEAKER_TAG = "speaker 1:"
SECOND_SPEAKER_TAG = "speaker 2:"

def get_speaker_positions(text, speaker_tag):

    total_interventions = text.count(speaker_tag)
    positions = []
    position = 0
    for i in range(total_interventions):
        positions.append(text.find(speaker_tag, position))
        # we increase the position by the addition of all the previous 
        # positions to reach the following occurrences through the list of 
        # positions
    position += sum(positions) + 1

    return positions

def slices(iterable, n):
    return zip(*(islice(it, i, None) for i, it in enumerate(tee(iterable, n))))

def get_text_interventions(text, speaker_tags):

    # speakers' interventions of the text
    interventions = { speaker_tag: "" for speaker_tag in speaker_tags }

    # positions where start each intervention in the text
    # (the last one is used to get the rest of the text, because it's the 
    # last intervention)
    # (we need to sort the positions to get the interventions in the correct 
    # order)
    speaker_positions = [
        get_speaker_positions(text, speaker) for speaker in speaker_tags
    ]
    all_positions = [
        position for sublist in speaker_positions for position in sublist
    ]
    all_positions.append(len(text))
    all_positions.sort()

    # generate the list of pairs that match a certain intervention
    # the pairs are formed by the initial and the end position of the 
    # intervention
    text_chunks = list(slices(all_positions, 2))

    for chunk in text_chunks:

        # we assign the intervention according to which 
        # list of speaker interventions the position exists
        # when slicing we add the speaker tag's length to exclude 
        # the speaker tag from the own intervention
        if chunk[0] in speaker_positions[0]:
            intervention = text[chunk[0]+len(speaker_tags[0]):chunk[1]]
            interventions[speaker_tags[0]] += intervention

        elif chunk[0] in speaker_positions[1]:
            intervention = text[chunk[0]+len(speaker_tags[1]):chunk[1]]
            interventions[speaker_tags[1]] += intervention

    return interventions

text_interventions = get_text_interventions(text, [ FIRST_SPEAKER_TAG, SECOND_SPEAKER_TAG ])

Notes:

• The credits for the slices() function is owned to the following answer: Find consecutive combinations

If you have any doubt, you can read more details in the itertools documentation:

• Documentation about itertools.islice and itertools.tee: islice tee

Feel free to ask me anything you didn't understand about the example. I hope you find it useful! =)