Deep Syntax - Revision history

Tamchyna: /* Using Deep Syntax to Achieve State of the Art in MT */

2015-11-17T17:25:26Z

Using Deep Syntax to Achieve State of the Art in MT

← Older revision		Revision as of 17:25, 17 November 2015
Line 45:		Line 45:
	Chimera currently represents the state of the art in English-Czech MT; it was ranked first by human judges in three consecutive years of the WMT shared Translation Task (2013, 2014, 2015).		Chimera currently represents the state of the art in English-Czech MT; it was ranked first by human judges in three consecutive years of the WMT shared Translation Task (2013, 2014, 2015).

	The following table shows the improvements from adding extra data and from including the TectoMT outputs. The results suggest that TectoMT provides improvements which are complementary and which significantly help translation quality:		The following table shows the improvements from adding extra data and from including the TectoMT outputs. The results suggest that TectoMT provides improvements which are complementary and which significantly help translation quality. The constrained setup used only 15 million parallel sentence pairs, as opposed to the full system with over 52 million training sentence pairs. In terms of monolingual data, the difference was 44 vs. 392 million sentences.

	{\| class="wikitable"		{\| class="wikitable"

Tamchyna: /* Using Deep Syntax to Achieve State of the Art in MT */

2015-11-17T17:23:11Z

Using Deep Syntax to Achieve State of the Art in MT

← Older revision		Revision as of 17:23, 17 November 2015
Line 53:		Line 53:
	! Delta		! Delta
	\|-		\|-
	\|CH0		!CH0
	\|21.28		\|21.28
	\|22.29		\|22.29
	\|1.31		\|1.31
	\|-		\|-
	\|CH1		!CH1
	\|23.37		\|23.37
	\|24.24		\|24.24
	\|0.87		\|0.87
	\|-		\|-
	\|Delta		!Delta
	\|2.09		\|2.09
	\|1.65		\|1.65
	\|		\|
	\|}		\|}

Tamchyna: /* Using Deep Syntax to Achieve State of the Art in MT */

2015-11-17T17:22:39Z

Using Deep Syntax to Achieve State of the Art in MT

← Older revision		Revision as of 17:22, 17 November 2015
Line 47:		Line 47:
	The following table shows the improvements from adding extra data and from including the TectoMT outputs. The results suggest that TectoMT provides improvements which are complementary and which significantly help translation quality:		The following table shows the improvements from adding extra data and from including the TectoMT outputs. The results suggest that TectoMT provides improvements which are complementary and which significantly help translation quality:

	{\|		{\| class="wikitable"
	!		!
	! Constrained		! Constrained

Tamchyna: /* Using Deep Syntax to Achieve State of the Art in MT */

2015-11-17T17:21:38Z

Using Deep Syntax to Achieve State of the Art in MT

← Older revision		Revision as of 17:21, 17 November 2015
Line 46:		Line 46:

	The following table shows the improvements from adding extra data and from including the TectoMT outputs. The results suggest that TectoMT provides improvements which are complementary and which significantly help translation quality:		The following table shows the improvements from adding extra data and from including the TectoMT outputs. The results suggest that TectoMT provides improvements which are complementary and which significantly help translation quality:

			{\|
			!
			! Constrained
			! Full
			! Delta
			\|-
			\|CH0
			\|21.28
			\|22.29
			\|1.31
			\|-
			\|CH1
			\|23.37
			\|24.24
			\|0.87
			\|-
			\|Delta
			\|2.09
			\|1.65
			\|
			\|}

Tamchyna: /* Using Deep Syntax to Achieve State of the Art in MT */

2015-11-17T17:17:43Z

Using Deep Syntax to Achieve State of the Art in MT

← Older revision		Revision as of 17:17, 17 November 2015
Line 38:		Line 38:

	[http://ufal.mff.cuni.cz/chimera Chimera] is a system combination of a standard phrase-based MT system and TectoMT. Development and test data are translated using TectoMT and its outputs are added as a separate parallel corpus. An extra phrase table is extracted from this synthetic set and is added to Moses. The system can therefore choose to use either the standard parallel data or the outputs of TectoMT. Standard MERT is used to set the weights.		[http://ufal.mff.cuni.cz/chimera Chimera] is a system combination of a standard phrase-based MT system and TectoMT. Development and test data are translated using TectoMT and its outputs are added as a separate parallel corpus. An extra phrase table is extracted from this synthetic set and is added to Moses. The system can therefore choose to use either the standard parallel data or the outputs of TectoMT. Standard MERT is used to set the weights.

			The following figure illustrates the structure of Chimera -- the final system is CH2 but CH1 (Moses+TectoMT but no automatic post-editing) and CH0 (only plain Moses) have also been evaluated.

	[[File:chimera.png\|400px]]		[[File:chimera.png\|400px]]

			Chimera currently represents the state of the art in English-Czech MT; it was ranked first by human judges in three consecutive years of the WMT shared Translation Task (2013, 2014, 2015).

			The following table shows the improvements from adding extra data and from including the TectoMT outputs. The results suggest that TectoMT provides improvements which are complementary and which significantly help translation quality:

Tamchyna: /* Using Deep Syntax to Achieve State of the Art in MT */

2015-11-14T09:12:56Z

Using Deep Syntax to Achieve State of the Art in MT

← Older revision		Revision as of 09:12, 14 November 2015
Line 39:		Line 39:
	[http://ufal.mff.cuni.cz/chimera Chimera] is a system combination of a standard phrase-based MT system and TectoMT. Development and test data are translated using TectoMT and its outputs are added as a separate parallel corpus. An extra phrase table is extracted from this synthetic set and is added to Moses. The system can therefore choose to use either the standard parallel data or the outputs of TectoMT. Standard MERT is used to set the weights.		[http://ufal.mff.cuni.cz/chimera Chimera] is a system combination of a standard phrase-based MT system and TectoMT. Development and test data are translated using TectoMT and its outputs are added as a separate parallel corpus. An extra phrase table is extracted from this synthetic set and is added to Moses. The system can therefore choose to use either the standard parallel data or the outputs of TectoMT. Standard MERT is used to set the weights.

	[[File:chimera.png\|~~500px~~]]		[[File:chimera.png\|400px]]

Tamchyna at 09:12, 14 November 2015

2015-11-14T09:12:40Z

← Older revision		Revision as of 09:12, 14 November 2015
Line 39:		Line 39:
	[http://ufal.mff.cuni.cz/chimera Chimera] is a system combination of a standard phrase-based MT system and TectoMT. Development and test data are translated using TectoMT and its outputs are added as a separate parallel corpus. An extra phrase table is extracted from this synthetic set and is added to Moses. The system can therefore choose to use either the standard parallel data or the outputs of TectoMT. Standard MERT is used to set the weights.		[http://ufal.mff.cuni.cz/chimera Chimera] is a system combination of a standard phrase-based MT system and TectoMT. Development and test data are translated using TectoMT and its outputs are added as a separate parallel corpus. An extra phrase table is extracted from this synthetic set and is added to Moses. The system can therefore choose to use either the standard parallel data or the outputs of TectoMT. Standard MERT is used to set the weights.

	[[File:chimera.png]]		[[File:chimera.png\|500px]]

Tamchyna: /* Using Deep Syntax to Achieve State of the Art in MT */

2015-11-14T09:11:57Z

Using Deep Syntax to Achieve State of the Art in MT

← Older revision		Revision as of 09:11, 14 November 2015
Line 38:		Line 38:

	[http://ufal.mff.cuni.cz/chimera Chimera] is a system combination of a standard phrase-based MT system and TectoMT. Development and test data are translated using TectoMT and its outputs are added as a separate parallel corpus. An extra phrase table is extracted from this synthetic set and is added to Moses. The system can therefore choose to use either the standard parallel data or the outputs of TectoMT. Standard MERT is used to set the weights.		[http://ufal.mff.cuni.cz/chimera Chimera] is a system combination of a standard phrase-based MT system and TectoMT. Development and test data are translated using TectoMT and its outputs are added as a separate parallel corpus. An extra phrase table is extracted from this synthetic set and is added to Moses. The system can therefore choose to use either the standard parallel data or the outputs of TectoMT. Standard MERT is used to set the weights.

			[[File:chimera.png]]

Tamchyna at 08:53, 14 November 2015

2015-11-14T08:53:20Z

← Older revision		Revision as of 08:53, 14 November 2015
Line 32:		Line 32:

	[[File:tectomt-example.png\|800px]]		[[File:tectomt-example.png\|800px]]

			== Using Deep Syntax to Achieve State of the Art in MT ==

			By itself, deep syntactic MT does not reach the performance of statistical methods (e.g. phrase-based). However, the outputs of TectoMT are usually grammatical sentences (as they are generated from a deep representation, preserving agreement constraints) and they can contain word forms not observed in the training data (thanks to its morphological generator). As such, they are a useful complement of the statistical systems.

			[http://ufal.mff.cuni.cz/chimera Chimera] is a system combination of a standard phrase-based MT system and TectoMT. Development and test data are translated using TectoMT and its outputs are added as a separate parallel corpus. An extra phrase table is extracted from this synthetic set and is added to Moses. The system can therefore choose to use either the standard parallel data or the outputs of TectoMT. Standard MERT is used to set the weights.

Tamchyna at 14:28, 7 October 2015

2015-10-07T14:28:22Z

← Older revision		Revision as of 14:28, 7 October 2015
Line 31:		Line 31:
	The following picture shows an example of Czech-English translation.		The following picture shows an example of Czech-English translation.

	[[File:tectomt-example.png]]		[[File:tectomt-example.png\|800px]]