- bump `unstructured-inference` to `0.6.6`
- specify default model name for element detection to be
`detectron2_onnx` to keep current behavior
- NOTE: the updated inference package by default would use yolox as
element detection model; this will be evaluated and enabled in a
separated PR
---------
Co-authored-by: ryannikolaidis <1208590+ryannikolaidis@users.noreply.github.com>
Co-authored-by: badGarnet <badGarnet@users.noreply.github.com>
Closes GH Issue #1233.
### Summary
- add functionality to shrink all bounding boxes along x and y axes
(still centered around the same center point) before running xy-cut sort
### Evaluation
Run the followin gcommand for this
[PDF](https://utic-dev-tech-fixtures.s3.us-east-2.amazonaws.com/pastebin/patent-11723901-page2.pdf).
PYTHONPATH=. python examples/custom-layout-order/evaluate_xy_cut_sorting.py <file_path> <strategy>
### Summary
Uses `langdetect` to detect all languages present in the input document.
### Details
- Converts all language codes (whether user inputted or detected using
`langdetect`) to a standard ISO 639-3 code.
- Adds `languages` field to the metadata
- Will revisit how to nonstandardly represent simplified vs traditional
Chinese scripts internally (separate PR).
- Update ingest test results to add `languages` field to documents. Some
other side effects are changes in order of some elements and changes in
element categorization
### Test
You can test the detect_languages function individually by importing the
function and inputting a text sample and optionally a language:
```
text = "My lubimy mleko i chleb."
doc_langs = detect_languages(text)
print(doc_langs)
```
-> ['ces', 'pol', 'slk']
---------
Co-authored-by: Newel H <37004249+newelh@users.noreply.github.com>
Co-authored-by: ryannikolaidis <1208590+ryannikolaidis@users.noreply.github.com>
Co-authored-by: shreyanid <shreyanid@users.noreply.github.com>
Co-authored-by: Trevor Bossert <37596773+tabossert@users.noreply.github.com>
Co-authored-by: Ronny H <138828701+ron-unstructured@users.noreply.github.com>
**Summary**
Adds logic to combine broken numbered list for pdf fast strategy.
**Details**
Previously the document reads the numbered list items part of the
`layout-parser-paper-fast.pdf` file as:
```
'1. An off-the-shelf toolkit for applying DL models for layout detection, character'
'recognition, and other DIA tasks (Section 3)'
'2. A rich repository of pre-trained neural network models (Model Zoo) that'
'underlies the off-the-shelf usage'
'3. Comprehensive tools for efficient document image data annotation and model'
'tuning to support different levels of customization'
'4. A DL model hub and community platform for the easy sharing, distribu- tion, and discussion of DIA models and pipelines, to promote reusability, reproducibility, and extensibility (Section 4)'
```
Now it reads:
```
'1. An off-the-shelf toolkit for applying DL models for layout detection, character recognition, and other DIA tasks (Section 3)'
'2. A rich repository of pre-trained neural network models (Model Zoo) that underlies the off-the-shelf usage'
'3. Comprehensive tools for efficient document image data annotation and model' tuning to support different levels of customization'
'4. A DL model hub and community platform for the easy sharing, distribu- tion, and discussion of DIA models and pipelines, to promote reusability, reproducibility, and extensibility (Section 4)'
```
The added logic leverages `ElementType` and `coordinates` to determine
whether the following lines is a part of the previously detected
`ListItem` or not.
**Test**
Add test that checks the element length less than original version with
broken numbered list. The test also checks whether the first detected
numbered list ends with previously broken line.
---------
Co-authored-by: ryannikolaidis <1208590+ryannikolaidis@users.noreply.github.com>
Co-authored-by: Klaijan <Klaijan@users.noreply.github.com>
Currently there are some cases when `partition_pdf` is run using the
`hi_res` strategy, in which elements can come back with category
`UncategorizedText`. This happens when the detection model fails to
detect an element, but we're able to find it anyway either because it
was embedded in the PDF, or we found it using OCR.
This commit is to allow for attempting to categorize these uncategorized
elements using our text-based classification function,
`element_from_text`.
### Summary
Closes#1230. Updates `partition_html` to split on `<br>` tags that
appear within text elements.
### Testing
The following is code previously produced one giant element on `main`.
```python
from unstructured.partition.html import partition_html
filename = "example-docs/ideas-page.html"
elements = partition_html(filename=filename)
len(elements) # Should be 4
print("\n\n".join([str(el) for el in elements)])
```
The output should be:
```python
January 2023
(Someone fed my essays into GPT to make something that could answer
questions based on them, then asked it where good ideas come from. The
answer was ok, but not what I would have said. This is what I would have said.)
The way to get new ideas is to notice anomalies: what seems strange,
or missing, or broken? You can see anomalies in everyday life (much
of standup comedy is based on this), but the best place to look for
them is at the frontiers of knowledge.
Knowledge grows fractally.
From a distance its edges look smooth, but when you learn enough
to get close to one, you'll notice it's full of gaps. These gaps
will seem obvious; it will seem inexplicable that no one has tried
x or wondered about y. In the best case, exploring such gaps yields
whole new fractal buds.
```
The issue was that for blocks detected in an image such as:
, where the full image is:
https://utic-dev-tech-fixtures.s3.us-east-2.amazonaws.com/pastebin//Users/cragwolfe/tmp/IRS-form-1987.png
, many ListItem's would be extracted that were not adding much value to
the output (assuming the block was determined to be of type List from
the layout model). This particular file is also used in ingest tests,
and you can see the prior output here:
https://github.com/Unstructured-IO/unstructured/blob/483b09b/test_unstructured_ingest/expected-structured-output/azure/IRS-form-1987.png.json#L93-L280
Test Instructions:
1. run the following snippet:
```
import json
import os
from datetime import datetime
from unstructured.__version__ import __version__
from unstructured.partition.auto import partition
from unstructured.staging.base import elements_to_json
filename = "/opt/home/tmp/IRS-form-1987.png"
output_dir = "/opt/home/tmp/json"
base_name_with_ext = os.path.basename(filename)
output_filename_part = os.path.join(output_dir, base_name_with_ext)
print(f"unstructured version: {__version__}")
#for strategy in ("hi_res", "fast", "auto"):
for strategy in ("hi_res",):
d1 = datetime.now()
elements = partition(filename=filename, strategy=strategy)
elems_as_dicts = json.loads(elements_to_json(elements, indent=2))
# strip out metadata for the sake of more readable results
for element_dict in elems_as_dicts:
del element_dict["metadata"]
json_filename=f"{output_filename_part}-{strategy}.json"
with open(json_filename, "w") as jsonf:
jsonf.write(json.dumps(elems_as_dicts, indent=2))
d2 = datetime.now()
print(f"num elements for {strategy}: {len(elements)}")
print(f"time elapsed {strategy}: {(d2-d1).total_seconds()}")
```
updating the `filename` and `output_dir` paths for your particular local
environment.
2. Open the json file that was writen to your `output_dir`, named
IRS-form-1987.png-hi_res.json
Witness the new element:
```
{
"type": "ListItem",
"element_id": "7d3ba328af2c20ddeef5d2c1d270f60f",
"text": "Long-term contracts.\u2014If you are required to change your method of accounting for long-term contracts under section 460, see Notice 87
-61 (9/21/87), 1987-38 IRB 40, for the notification procedures that must be followed Other methods. \u2014Unless the Service has Published a regulation
or procedure to the contrary, all other changes in accounting methods required by the Act are automatically considered to be approved by the Commissio
ner. Examples of method changes automatically approved by the Commissioner are those changes required to effect: (1) the repeal of the reserve method f
or bad debts of taxpayers other than financial institutions (Act section 805); (2) the repeal of the installment method for sales under a revolving cre
dit plan (Act section 812); (3) the Inclusion of income attributable to the sale or furnishing of utility services no later than the year in which the
services were provided to customers (Act section 821); and (4) the repeal of the deduction for qualified discount coupons (Act section 823). Do not fil
e Form 3115 for these changes."
},
```
### Summary
Address
[#1136](https://github.com/Unstructured-IO/unstructured/issues/1136) for
`hi_res` and `fast` strategies. The `ocr_only` strategy does not include
coordinates.
- add functionality to switch sort mode between the current `basic`
sorting and the new `xy-cut` sorting for `hi_res` and `fast` strategies
- add the script to evaluate the `xy-cut` sorting approach
- add jupyter notebook to provide evaluation and visualization for the
`xy-cut` sorting approach
### Evaluation
```
export PYTHONPATH=.:$PYTHONPATH && python examples/custom-layout-order/evaluate_xy_cut_sorting.py <file_path> <strategy>
```
Here, the file should be under the project root directory. For example,
```
export PYTHONPATH=.:$PYTHONPATH && python examples/custom-layout-order/evaluate_xy_cut_sorting.py example-docs/multi-column-2p.pdf fast
```
* pip-compile in order to bump unstructured-inference
* Set the default `ocr_mode` back to `enitre_page` now that [this
error](https://github.com/Unstructured-IO/unstructured-inference/pull/183)
is addressed
* Explicitly add `sphinx-tabs` to `build.in`. This file provides
`docs/requirements.txt`.
* Remove a pinned `pydantic` version
* Fix a makefile command to `pip-compile` a missing ingest file.
Set to individual_blocks for now to work around [this
bug](https://github.com/Unstructured-IO/unstructured-inference/issues/179).
I verified by printing the current ocr_mode in inference. The
`entire_page` default is overridden.
---------
Co-authored-by: ryannikolaidis <1208590+ryannikolaidis@users.noreply.github.com>
Co-authored-by: awalker4 <awalker4@users.noreply.github.com>
Bump to unstructured-inference==0.5.13, which includes:
Fix extracted image elements being included in layout merge, addresses the issue
where an entire-page image in a PDF was not passed to the layout model when using hi_res.
* track tags in html
* pass through links as metadata
* add test for grabbing links
* one more link
* changelog and version
* update docs
* fix tests
* update empty link assertion
* ingest-test-fixtures-update
* Update ingest test fixtures (#961)
More deterministic element ordering when using hi_res PDF parsing strategy (from unstructured-inference bump to 0.5.4)
Make large model available (from unstructured-inference bump to 0.5.3)
Combine inferred elements with extracted elements (from unstructured-inference bump to 0.5.2)
---------
Co-authored-by: Roman Isecke <roman@unstructured.io>
Co-authored-by: Crag Wolfe <crag@unstructured.io>
- Adds reusable validation scripts (check-x.sh) to minimize repeated (or near-repeated) code and create one source of truth
- Restructures the location of download and output folders such that they are nested in the test_unstructured_ingest directory
- Adds gitignore for output folders / files to avoid them accidentally getting checked into the repository
- Construct paths as reusable variables declared at top of scripts
- Sort order of flag for ingest calls, across all tests (this makes it easier to parse at a glance)
- OVERWRITE_FIXTURES removes all old fixtures for path to guarantee no stale results are left behind
- Bonus: don't check/exit on expected number of expected outputs when OVERWRITE_FIXTURES is true
- Bonus: exclude file_directory from Slack and Discord test scripts (match convention in all others)
