-
### Description
This is for the Automating the MSE project.
Clinical Observation | Observation Sub-Components | Potential Features + Approaches
-- | -- | --
Speed | | Mean length of clause, T-…
-
This seems wrong: why does we get such a huge surprisal for sentence-end after a period? Input file was:
```
This is a short sentence.
```
Command & output:
```
$ lm-zoo get-surprisals ng…
rlevy updated
3 years ago
-
Sass et al. (2012) found an interesting pattern in valence priming: positive words are highly effective in priming other positive words, but negative words don't really "prime" anything (that is, RT o…
-
```python
Traceback (most recent call last):
File "/net/vast-storage/scratch/vast/evlab/asathe/code/composlang/lmsurprisal/notebooks/extract_surprisals.py", line 73, in
main()
File "/net/…
-
To do:
- [x] Get relevant n-grams of the corpora.
- [ ] Compare different n-grams for co-occurrence in both English and US corpora.
- [ ] Check out surprisal tool - used to be in NLTK. Find out why …
-
It would be great if `syntaxgym evaluate` had an option for returning by-region surprisals in a tabular format (something easily readable as a data frame in `R` or `pandas`). Probably the most natural…
rlevy updated
3 years ago
-
Hi,
Minicons seems not tokenize alphabet-based texts given "bert" pre-trained models are introduced. It is desirable to generate the surprisal values for word forms in real life rather than the sp…
-
I am working on replicating the analyses/code from the repo and I wanted to point out that the data are encoded for some items as "None" being the first word. Pandas (as coded in the scripts) will aut…
-
From GB:
> Test whether post-error reading (accuracy and/or rt) differs in the context of a negative vs positive passage, with the rationale being that errors are negatively valenced and so there is a…
-
The `vocab_size` in the `config.json` is set to 52000, however, the `eos_token_id` is also set to 52000, which causes an `IndexError` in the forward pass if you have an eos token in your `input_ids`.
…