HU TN Fixes issue #166

Jenkinsfile

@@ -17,7 +17,7 @@ pipeline {
     ES_TN_CACHE='/home/jenkinsci/TestData/text_norm/ci/grammars/10-27-23-0'
     ES_EN_TN_CACHE='/home/jenkinsci/TestData/text_norm/ci/grammars/06-13-23-2'
     FR_TN_CACHE='/home/jenkinsci/TestData/text_norm/ci/grammars/06-04-24-0'
-    HU_TN_CACHE='/home/jenkinsci/TestData/text_norm/ci/grammars/06-08-23-0'
+    HU_TN_CACHE='/home/jenkinsci/TestData/text_norm/ci/grammars/06-14-24-0'
     PT_TN_CACHE='/home/jenkinsci/TestData/text_norm/ci/grammars/06-08-23-0'
     RU_TN_CACHE='/home/jenkinsci/TestData/text_norm/ci/grammars/06-08-23-0'
     VI_TN_CACHE='/home/jenkinsci/TestData/text_norm/ci/grammars/06-08-23-0'

nemo_text_processing/text_normalization/hu/taggers/electronic.py

@@ -11,14 +11,13 @@
 # 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.
+
+
 import pynini
 from pynini.lib import pynutil
 
+from nemo_text_processing.text_normalization.de.utils import get_abs_path, load_labels
 from nemo_text_processing.text_normalization.en.graph_utils import NEMO_ALPHA, NEMO_DIGIT, GraphFst, insert_space
-from nemo_text_processing.text_normalization.es.utils import get_abs_path, load_labels
-
-common_domains = [x[0] for x in load_labels(get_abs_path("data/electronic/domain.tsv"))]
-symbols = [x[0] for x in load_labels(get_abs_path("data/electronic/symbols.tsv"))]
 
 
 class ElectronicFst(GraphFst):
@@ -36,28 +35,35 @@ def __init__(self, deterministic: bool = True):
         super().__init__(name="electronic", kind="classify", deterministic=deterministic)
 
         dot = pynini.accep(".")
-        accepted_common_domains = pynini.union(*common_domains)
-        accepted_symbols = pynini.union(*symbols) - dot
-        accepted_characters = pynini.closure(NEMO_ALPHA | NEMO_DIGIT | accepted_symbols)
-        acceepted_characters_with_dot = pynini.closure(NEMO_ALPHA | NEMO_DIGIT | accepted_symbols | dot)
 
-        # email
-        username = (
-            pynutil.insert("username: \"")
-            + acceepted_characters_with_dot
-            + pynutil.insert("\"")
-            + pynini.cross('@', ' ')
+        symbols = [x[0] for x in load_labels(get_abs_path("data/electronic/symbols.tsv"))]
+        symbols = pynini.union(*symbols)
+        # all symbols
+        symbols_no_period = pynini.difference(symbols, dot)  # alphabet of accepted symbols excluding the '.'
+        accepted_characters = pynini.closure(
+            (NEMO_ALPHA | NEMO_DIGIT | symbols_no_period), 1
+        )  # alphabet of accepted chars excluding the '.'
+        all_characters = pynini.closure(
+            (NEMO_ALPHA | NEMO_DIGIT | symbols), 1
+        )  # alphabet of accepted chars including the '.'
+
+        # domains
+        domain = dot + accepted_characters
+        domain_graph = (
+            pynutil.insert('domain: "') + (accepted_characters + pynini.closure(domain, 1)) + pynutil.insert('"')
         )
-        domain_graph = accepted_characters + dot + accepted_characters
-        domain_graph = pynutil.insert("domain: \"") + domain_graph + pynutil.insert("\"")
-        domain_common_graph = (
-            pynutil.insert("domain: \"")
-            + accepted_characters
-            + accepted_common_domains
-            + pynini.closure((accepted_symbols | dot) + pynini.closure(accepted_characters, 1), 0, 1)
-            + pynutil.insert("\"")
+
+        # email
+        username = pynutil.insert('username: "') + all_characters + pynutil.insert('"') + pynini.cross("@", " ")
+        email = username + domain_graph
+
+        # social media tags
+        tag = (
+            pynini.cross("@", "")
+            + pynutil.insert('username: "')
+            + (accepted_characters | (accepted_characters + pynini.closure(domain, 1)))
+            + pynutil.insert('"')
         )
-        graph = (username + domain_graph) | domain_common_graph
 
         # url
         protocol_start = pynini.accep("https://") | pynini.accep("http://")
@@ -72,8 +78,10 @@ def __init__(self, deterministic: bool = True):
             )
         )
         protocol = protocol_start | protocol_end | (protocol_start + protocol_end)
-        protocol = pynutil.insert("protocol: \"") + protocol + pynutil.insert("\"")
-        graph |= protocol + insert_space + (domain_graph | domain_common_graph)
+        protocol = pynutil.insert('protocol: "') + protocol + pynutil.insert('"')
+        url = protocol + insert_space + (domain_graph)
+
+        graph = url | domain_graph | email | tag
         self.graph = graph
 
         final_graph = self.add_tokens(self.graph + pynutil.insert(" preserve_order: true"))

nemo_text_processing/text_normalization/hu/taggers/tokenize_and_classify.py

@@ -69,7 +69,7 @@ def __init__(
             os.makedirs(cache_dir, exist_ok=True)
             whitelist_file = os.path.basename(whitelist) if whitelist else ""
             far_file = os.path.join(
-                cache_dir, f"_{input_case}_hu_tn_{deterministic}_deterministic{whitelist_file}.far"
+                cache_dir, f"_{input_case}_hu_tn_{deterministic}_deterministic{whitelist_file}.far",
             )
         if not overwrite_cache and far_file and os.path.exists(far_file):
             self.fst = pynini.Far(far_file, mode="r")["tokenize_and_classify"]
@@ -86,10 +86,10 @@ def __init__(
             self.decimal = DecimalFst(cardinal=self.cardinal, deterministic=deterministic)
             decimal_graph = self.decimal.fst
 
-            self.fraction = FractionFst(cardinal=self.cardinal, ordinal=self.ordinal, deterministic=deterministic)
+            self.fraction = FractionFst(cardinal=self.cardinal, ordinal=self.ordinal, deterministic=deterministic,)
             fraction_graph = self.fraction.fst
             self.measure = MeasureFst(
-                cardinal=self.cardinal, decimal=self.decimal, fraction=self.fraction, deterministic=deterministic
+                cardinal=self.cardinal, decimal=self.decimal, fraction=self.fraction, deterministic=deterministic,
             )
             measure_graph = self.measure.fst
             self.date = DateFst(cardinal=self.cardinal, deterministic=deterministic)
@@ -101,7 +101,7 @@ def __init__(
             telephone_graph = self.telephone.fst
             self.electronic = ElectronicFst(deterministic=deterministic)
             electronic_graph = self.electronic.fst
-            self.money = MoneyFst(cardinal=self.cardinal, decimal=self.decimal, deterministic=deterministic)
+            self.money = MoneyFst(cardinal=self.cardinal, decimal=self.decimal, deterministic=deterministic,)
             money_graph = self.money.fst
             self.whitelist = WhiteListFst(input_case=input_case, deterministic=deterministic, input_file=whitelist)
             whitelist_graph = self.whitelist.fst
@@ -118,7 +118,7 @@ def __init__(
                 | pynutil.add_weight(decimal_graph, 1.1)
                 | pynutil.add_weight(money_graph, 1.1)
                 | pynutil.add_weight(telephone_graph, 1.1)
-                | pynutil.add_weight(electronic_graph, 1.1)
+                | pynutil.add_weight(electronic_graph, 1.11)
                 | pynutil.add_weight(word_graph, 200)
             )
             punct = pynutil.insert("tokens { ") + pynutil.add_weight(punct_graph, weight=2.1) + pynutil.insert(" }")

nemo_text_processing/text_normalization/hu/verbalizers/electronic.py

@@ -17,6 +17,7 @@
 from nemo_text_processing.text_normalization.en.graph_utils import (
     NEMO_NOT_QUOTE,
     NEMO_SIGMA,
+    NEMO_SPACE,
     GraphFst,
     delete_preserve_order,
     insert_space,
@@ -64,21 +65,24 @@ def add_space_after_char():
 
         verbalize_characters = pynini.cdrewrite(graph_symbols | graph_digit, "", "", NEMO_SIGMA)
 
-        user_name = pynutil.delete("username: \"") + add_space_after_char() + pynutil.delete("\"")
+        user_name = pynutil.delete('username: "') + add_space_after_char() + pynutil.delete('"')
         user_name @= verbalize_characters
 
         convert_defaults = pynutil.add_weight(NEMO_NOT_QUOTE, weight=0.0001) | domain_common | server_common
         domain = convert_defaults + pynini.closure(insert_space + convert_defaults)
         domain @= verbalize_characters
 
-        domain = pynutil.delete("domain: \"") + domain + pynutil.delete("\"")
+        domain = pynutil.delete('domain: "') + domain + pynutil.delete('"')
         protocol = (
-            pynutil.delete("protocol: \"")
+            pynutil.delete('protocol: "')
             + add_space_after_char() @ pynini.cdrewrite(graph_symbols, "", "", NEMO_SIGMA)
-            + pynutil.delete("\"")
+            + pynutil.delete('"')
         )
-        self.graph = (pynini.closure(protocol + pynini.accep(" "), 0, 1) + domain) | (
-            user_name + pynini.accep(" ") + at_sign + domain
+        # self.graph = (pynini.closure(protocol + pynini.accep(" "), 0, 1) + domain) | (
+        #     user_name + pynini.accep(" ") + at_sign + domain
+        # )
+        self.graph = (pynini.closure(protocol + NEMO_SPACE, 0, 1) + domain) | (
+            user_name + NEMO_SPACE + pynutil.insert("kukac ") + domain | (pynutil.insert("kukac ") + user_name)
         )
         delete_tokens = self.delete_tokens(self.graph + delete_preserve_order)
         self.fst = delete_tokens.optimize()

tests/nemo_text_processing/hu/data_text_normalization/test_cases_electronic.txt

@@ -1,2 +1,8 @@
 abc@def.hu~a b c kukac d e f pont hu
 https://www.nvidia.com~h t t p s kettőspont perjel perjel w w w pont nvidia pont com
+https://www.nvidia.com.~h t t p s kettőspont perjel perjel w w w pont nvidia pont com .
+@jensen~kukac j e n s e n
+www.nasa.gov.~w w w pont n a s a pont gov .
+www.enveedya.com.hu.~w w w pont e n v e e d y a pont com pont hu .
+@jensen.me~kukac j e n s e n pont m e
+@wezyr1986~kukac w e z y r egy kilenc nyolc hat

tests/nemo_text_processing/hu/test_sparrowhawk_normalization.sh

@@ -1,7 +1,7 @@
 #! /bin/sh
 
 GRAMMARS_DIR=${1:-"/workspace/sparrowhawk/documentation/grammars"}
-PROJECT_DIR=${2:-"/workspace/tests/en"}
+PROJECT_DIR=${2:-"/workspace/tests/"}
 
 runtest () {
   input=$1