Extract text from DAT

2025-08-18 14:00 UTC gpt-5-thinking Open in ChatGPT ↗

Easiest way:

Make a Text DAT (e.g., latest_text).
In its contents, put a TD expression that reads that one cell:

`op('transcript_chunks')[0,2].val`

That will keep the Text DAT always equal to Row 0, Col 2 of your table DAT (transcript_chunks).

If you prefer to “push” the update when the table changes:

Add a DAT Execute DAT and set its DAT parameter to your table.
Enable onTableChange (or onCellChange).
In the callback:

def onTableChange(dat):
    op('latest_text').text = dat[0,2].val
    return

Either method will give you the latest text from row 0, column 2 in a Text DAT.

Nice — glad that part is working!
If you only want to grab a keyword / primary noun from the latest line, you’ll need a little bit of natural-language filtering inside your DAT Execute.

Since TouchDesigner runs Python, you can use a very lightweight approach:

Option 1: Just pick the longest word (quick + dirty)

def onTableChange(dat):
    text = dat[0,2].val   # latest text
    words = [w.strip(",.!?-") for w in text.split()]
    keyword = max(words, key=len) if words else ""
    op('latest_text').text = keyword
    return

Option 2: Use a noun extractor (better quality)

TouchDesigner ships with standard Python, so you can install nltk (or spacy) if you’re comfortable adding libraries. Example with nltk:

import nltk
from nltk import word_tokenize, pos_tag

def onTableChange(dat):
    text = dat[0,2].val
    tokens = word_tokenize(text)
    tagged = pos_tag(tokens)
    # Grab first noun
    nouns = [word for word,pos in tagged if pos.startswith('NN')]
    keyword = nouns[0] if nouns else ""
    op('latest_text').text = keyword
    return

⚡ If you want fast and simple inside TD without external installs, go with Option 1 or a rule-based filter (e.g., pick any word from a list you care about like "flowers", "sea", "boat").

Do you want me to write you a rule-based filter (so you can just define your own list of “primary nouns” to look for), or would you like to try hooking in nltk/spacy?

Got it 👍 — in TouchDesigner, almost any parameter can be filled dynamically with an expression that points to another operator’s value.

Since your latest_text is a Text DAT, the trick is to grab its .text contents.

For the Prompt parameter field, type an expression like:

op('latest_text').text

That will return the entire contents of the Text DAT as the parameter value.

✨ A couple of notes:

If you only want the first line of the Text DAT:
```
op('latest_text')[0,0]
```
(That’s row 0, col 0 of the DAT.)
If you want the whole DAT as one string (which is usually what you want for prompts):
```
op('latest_text').text
```

👉 Would you like the prompt parameter to continuously update live as new text comes in, or just be set once when you trigger something?

Perfect — if you just want the last word in that cell, you can change your function like this:

def onTableChange(dat):
    text = dat[1,2].val   # latest text
    words = [w.strip(",.!?-") for w in text.split()]
    keyword = words[-1] if words else ""
    op('latest_text').text = keyword
    return

That way, whenever the table updates, it grabs the last word in the row 1, col 2 cell, strips punctuation, and writes it to latest_text.

👉 Do you want it to grab the last word from row 0 (the newest chunk) instead of row 1? In your earlier screenshot, row 0 looked like the newest line.