Implement new end-to-end parser class

.github/workflows/build_test.yml

@@ -63,13 +63,22 @@ jobs:
       run: pip install .[extras] .[test]
     - name: Locate bobcat pre-trained model cache
       id: loc-bobcat-cache
-      run: echo "dir=$(python -c 'from lambeq.text2diagram.model_downloader import ModelDownloader; print(ModelDownloader("bert").model_dir)')" >> $GITHUB_OUTPUT
+      run: echo "dir=$(python -c 'from lambeq.text2diagram.model_based_reader.model_downloader import ModelDownloader; print(ModelDownloader("bobcat").model_dir)')" >> $GITHUB_OUTPUT
     - name: Restore bobcat pre-trained model from cache
       id: bobcat-cache
       uses: actions/cache@v4
       with:
         path: ${{ steps.loc-bobcat-cache.outputs.dir }}
-        key: bobcat-bert-v1
+        key: bobcat-v1
+    - name: Locate oncilla pre-trained model cache
+      id: loc-oncilla-cache
+      run: echo "dir=$(python -c 'from lambeq.text2diagram.model_based_reader.model_downloader import ModelDownloader; print(ModelDownloader("oncilla").model_dir)')" >> $GITHUB_OUTPUT
+    - name: Restore oncilla pre-trained model from cache
+      id: oncilla-cache
+      uses: actions/cache@v4
+      with:
+        path: ${{ steps.loc-oncilla-cache.outputs.dir }}
+        key: oncilla-v1
     - name: Test with pytest
       run: >
         coverage run --source=${{ env.SRC_DIR }}

lambeq/__init__.py

@@ -48,6 +48,8 @@
         'CCGBankParser',
         'DepCCGParseError',
         'DepCCGParser',
+        'OncillaParseError',
+        'OncillaParser',
         'WebParseError',
         'WebParser',
 
@@ -122,6 +124,7 @@
         CCGType, CCGRule, CCGRuleUseError, CCGTree,
         CCGParser,
         BobcatParseError, BobcatParser,
+        OncillaParseError, OncillaParser,
         CCGBankParseError, CCGBankParser,
         DepCCGParseError, DepCCGParser,
         WebParseError, WebParser,

lambeq/backend/__init__.py

@@ -25,8 +25,6 @@
            'Ty',
            'Word',
 
-           'PregroupTreeNode',
-
            'draw',
            'draw_equation',
            'to_gif',
@@ -35,6 +33,5 @@
 
 from lambeq.backend.grammar import (Box, Cap, Category, Cup, Diagram,
                                     Frame, Functor, Id, Spider, Swap, Ty, Word)
-from lambeq.backend.pregroup_tree import PregroupTreeNode
 from lambeq.backend.symbol import lambdify, Symbol
 from lambeq.backend.drawing import draw, draw_equation, to_gif

lambeq/backend/grammar.py

@@ -36,7 +36,7 @@
 
 if TYPE_CHECKING:
     import discopy
-    from lambeq.backend.pregroup_tree import PregroupTreeNode
+    from lambeq.text2diagram.pregroup_tree import PregroupTreeNode
 
 
 @dataclass

lambeq/bobcat/tagger.py

@@ -11,6 +11,15 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+"""
+CCG tagger model
+================
+
+Model for tagging text with CCG tags. This work is based on
+the PyTorch BERT model available in Huggingface transformers
+(https://huggingface.co/transformers) which is released under
+Apache License 2.0.
+"""
 
 from __future__ import annotations
 

lambeq/cli.py

@@ -39,12 +39,12 @@
 from lambeq.backend import grammar, tensor
 from lambeq.rewrite import RemoveSwapsRewriter
 from lambeq.text2diagram.base import Reader
-from lambeq.text2diagram.bobcat_parser import BobcatParser
 from lambeq.text2diagram.ccg_parser import CCGParser
 from lambeq.text2diagram.ccg_tree import CCGTree
 from lambeq.text2diagram.depccg_parser import DepCCGParser
 from lambeq.text2diagram.linear_reader import (cups_reader,
                                                stairs_reader)
+from lambeq.text2diagram.model_based_reader import BobcatParser
 from lambeq.text2diagram.spiders_reader import spiders_reader
 from lambeq.text2diagram.tree_reader import TreeReader
 from lambeq.tokeniser import SpacyTokeniser

lambeq/core/utils.py

@@ -19,8 +19,10 @@
 from typing import Any, List, Union
 
 
-SentenceType = Union[str, List[str]]
-SentenceBatchType = Union[List[str], List[List[str]]]
+TokenisedSentenceType = List[str]
+SentenceType = Union[str, TokenisedSentenceType]
+TokenisedSentenceBatchType = List[TokenisedSentenceType]
+SentenceBatchType = Union[List[str], TokenisedSentenceBatchType]
 
 
 def tokenised_sentence_type_check(sentence: SentenceType) -> bool:

lambeq/experimental/discocirc/reader.py

@@ -21,13 +21,13 @@
 
 from lambeq import AtomicType
 from lambeq.backend.grammar import Box, Diagram, Frame, Id, Spider, Ty
-from lambeq.backend.pregroup_tree import PregroupTreeNode
 from lambeq.core.utils import SentenceBatchType, SentenceType
 from lambeq.experimental.discocirc import (CoreferenceResolver,
                                            SpacyCoreferenceResolver,
                                            TreeRewriter,
                                            TreeRewriteRule)
 from lambeq.text2diagram import BobcatParser, CCGParser, Reader
+from lambeq.text2diagram.pregroup_tree import PregroupTreeNode
 
 
 NOUN = AtomicType.NOUN

lambeq/oncilla/__init__.py

@@ -0,0 +1,11 @@
+from lambeq.oncilla.parser import (
+    BertForSentenceToTree,
+    prepare_parent_logits_mask,
+    SentenceToTreeBertConfig,
+)
+
+__all__ = [
+    'BertForSentenceToTree',
+    'prepare_parent_logits_mask',
+    'SentenceToTreeBertConfig',
+]