Presentation is loading. Please wait.

Presentation is loading. Please wait.

Large Scale Machine Translation Architectures Qin Gao.

Similar presentations


Presentation on theme: "Large Scale Machine Translation Architectures Qin Gao."— Presentation transcript:

1 Large Scale Machine Translation Architectures Qin Gao

2 Outline Typical Problems in Machine Translation Program Model for Machine Translation MapReduce Required System Component Supporting software Distributed streaming data storage system Distributed structured data storage system Integrating – How to make a full-distributed system 2015-10-252Qin Gao, LTI, CMU

3 Why large scale MT We need more data.. But… 2015-10-253Qin Gao, LTI, CMU

4 Some representative MT problems Counting events in corpora ◦  Ngram count Sorting ◦  Phrase table extraction Preprocessing Data ◦  Parsing, tokenizing, etc Iterative optimization ◦  GIZA++ (All EM algorithms) 2015-10-254Qin Gao, LTI, CMU

5 Characteristics of different tasks Counting events in corpora ◦ Extract knowledge from data Sorting ◦ Process data, knowledge is inside data Preprocessing Data ◦ Process data, require external knowledge Iterative optimization ◦ For each iteration, process data using existing knowledge and update knowledge 2015-10-255Qin Gao, LTI, CMU

6 Components required for large scale MT Data Knowledge 2015-10-256Qin Gao, LTI, CMU

7 Components required for large scale MT Data Knowledge 2015-10-257Qin Gao, LTI, CMU

8 Components required for large scale MT Data Knowledge Stream Data Structured Knowledge Processor 2015-10-258Qin Gao, LTI, CMU

9 Problem for each component Stream data: ◦ As the amount of data grows, even a complete navigation is impossible. Processor: ◦ Single processor’s computation power is not enough Knowledge: ◦ The size of the table is too large to fit into memory ◦ Cache-based/distributed knowledge base suffers from low speed 2015-10-259Qin Gao, LTI, CMU

10 Make it simple: What is the underlying problem? We have a huge cake and we want to cut them into pieces and eat. Different cases: ◦ We just need to eat the cake. ◦ We also want to count how many peanuts inside the cake ◦ (Sometimes)We have only one folk! 2015-10-2510Qin Gao, LTI, CMU

11 Parallelization Data Knowledge 2015-10-2511Qin Gao, LTI, CMU

12 Solutions Large-scale distributed processing ◦ MapReduce: Simplified Data Processing on Large Clusters, Jeffrey Dean, Sanjay Ghemawat, Communications of the ACM, vol. 51, no. 1 (2008), pp. 107-113. Handling huge streaming data ◦ The Google File System, Sanjay Ghemawat, Howard Gobioff, Shun-Tak Leung, Proceedings of the 19th ACM Symposium on Operating Systems Principles, 2003, pp. 20-43. Handling structured data ◦ Large Language Models in Machine Translation, Thorsten Brants, Ashok C. Popat, Peng Xu, Franz J. Och, Jeffrey Dean, Proceedings of the 2007 Joint Conference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning (EMNLP-CoNLL), pp. 858-867. ◦ Bigtable: A Distributed Storage System for Structured Data, Fay Chang, Jeffrey Dean, Sanjay Ghemawat, Wilson C. Hsieh, Deborah A. Wallach, Mike Burrows, Tushar Chandra, Andrew Fikes, Robert E. Gruber, 7th USENIX Symposium on Operating Systems Design and Implementation (OSDI), 2006, pp. 205-218. 2015-10-2512Qin Gao, LTI, CMU

13 MapReduce MapReduce can refer to ◦ A programming model that deal with massive, unordered, streaming data processing tasks(MUD) ◦ A set of supporting software environment implemented by Google Inc Alternative implementation: ◦ Hadoop by Apache fundation 2015-10-2513Qin Gao, LTI, CMU

14 MapReduce programming model Abstracts the computation into two functions: ◦ MAP ◦ Reduce User is responsible for the implementation of the Map and Reduce functions, and supporting software take care of executing them 2015-10-2514Qin Gao, LTI, CMU

15 Representation of data The streaming data is abstracted as a sequence of key/value pairs Example: ◦ (sentence_id : sentence_content) 2015-10-2515Qin Gao, LTI, CMU

16 Map function The Map function takes an input key/value pair, and output a set of intermediate key/value pairs Key 1 : Value 1 Key 2 : Value 2 Map() Key 1 : Value 1 Key 2 : Value 2 Key 3 : Value 3 …….. Map() Key 1 : Value 2 Key 2 : Value 1 Key 3 : Value 3 …….. 2015-10-2516Qin Gao, LTI, CMU

17 Reduce function Reduce function accepts one intermediate key and a set of intermediate values, and produce the result Key 1 : Value 1 Key 1 : Value 2 Key 1 : Value 3 …….. Key 2 : Value 1 Key 2 : Value 2 Key 2 : Value 3 …….. Reduce() Result 2015-10-2517Qin Gao, LTI, CMU

18 The architecture of MapReduce Map function Reduce Function Distributed Sort 2015-10-2518Qin Gao, LTI, CMU

19 Benefit of MapReduce Automatic splitting data Fault tolerance High-throughput computing, uses the nodes efficiently Most important: Simplicity, just need to convert your algorithm to the MapReduce model. 2015-10-2519Qin Gao, LTI, CMU

20 Requirement for expressing algorithm in MapReduce Process Unordered data ◦ The data must be unordered, which means no matter in what order the data is processed, the result should be the same Produce Independent intermediate key ◦ Reduce function can not see the value of other keys 2015-10-2520Qin Gao, LTI, CMU

21 Example Distributed Word Count (1) ◦ Input key : word ◦ Input value : 1 ◦ Intermediate key : constant ◦ Intermediate value: 1 ◦ Reduce() : Count all intermediate values Distributed Word Count (2) ◦ Input key : Document/Sentence ID ◦ Input value : Document/Sentence content ◦ Intermediate key : constant ◦ Intermediate value: number of words in the document/sentence ◦ Reduce() : Count all intermediate values 2015-10-2521Qin Gao, LTI, CMU

22 Example 2 Distributed unigram count ◦ Input key : Document/Sentence ID ◦ Input value : Document/Sentence content ◦ Intermediate key : Word ◦ Intermediate value: Number of the word in the document/sentence ◦ Reduce() : Count all intermediate values 2015-10-2522Qin Gao, LTI, CMU

23 Example 3 Distributed Sort ◦ Input key : Entry key ◦ Input value : Entry content ◦ Intermediate key : Entry key (modification may be needed for ascend/descend order) ◦ Intermediate value: Entry content ◦ Reduce() : All the entry content Making use of built-in sorting functionality 2015-10-2523Qin Gao, LTI, CMU

24 Supporting MapReduce: Distributed Storage Reminder what we are dealing with in MapReduce: ◦ Massive, unordered, streaming data Motivation: ◦ We need to store large amount of data ◦ Make use of storage in all the nodes ◦ Automatic replication  Fault tolerant  Avoid hot spots client can read from many servers Google FS and Hadoop FS (HDFS) 2015-10-2524Qin Gao, LTI, CMU

25 Design principle of Google FS Optimizing for special workload: ◦ Large streaming reads, small random reads ◦ Large streaming writes, rare modification Support concurrent appending ◦ It actually assumes data are unordered High sustained bandwidth is more important than low latency, fast response time is not important Fault tolerant 2015-10-2525Qin Gao, LTI, CMU

26 Google FS Architecture Optimize for large streaming reading and large, concurrent writing Small random reading/writing is also supported, but not optimized Allow appending to existing files File are spitted into chunks and stored in several chunk servers A master is responsible for storage and query of chunk information 2015-10-2526Qin Gao, LTI, CMU

27 Google FS architecture 2015-10-2527Qin Gao, LTI, CMU

28 Replication When a chunk is frequently or “simultaneously” read from a client, the client may fail A fault in one client may cause the file not usable Solution: store the chunks in multiple machines. The number of replica of each chunk : replication factor 2015-10-2528Qin Gao, LTI, CMU

29 HDFS HDFS shares similar design principle of Google FS Write-once-read-many : Can only write file once, even appending is now allowed “Moving computation is cheaper than moving data” 2015-10-2529Qin Gao, LTI, CMU

30 Are we done? NO… Problems about the existing architecture 2015-10-2530Qin Gao, LTI, CMU

31 We are good at dealing with data What about knowledge? I.E. structured data? What if the size of the knowledge is HUGE? 2015-10-2531Qin Gao, LTI, CMU

32 A good example: GIZA A typical EM algorithm World Alignment Collect Counts Has More Sentences? Y Normalize Counts N Has More Iterations? Y N 2015-10-2532Qin Gao, LTI, CMU

33 When parallelized: seems to be a perfect MapReduce application Word Alignment Collect Counts Has More Sentences? Y Normalize Counts N Has More Iterations? Y N Word Alignment Collect Counts Has More Sentences? Y Word Alignment Collect Counts Has More Sentences? Y NN Run on cluster 2015-10-2533Qin Gao, LTI, CMU

34 However: ….................................................................... Large parallel corpus Corpus chunks Count tables Combined count table Statistical lexicon Renormalization Redistribute for next iteration Memory Data I/O Map Reduce Memory 2015-10-2534Qin Gao, LTI, CMU

35 Huge tables Lexicon probability table: T-Table Up to 3G in early stages As the number of workers increases, they all need to load this 3G file! And all the nodes need to have 3G+ memory – we need a cluster of super computers? 2015-10-2535Qin Gao, LTI, CMU

36 Another example, decoding Consider language models, what can we do if the language model grows to several TBs We need storage/query mechanism for large, structured data Consideration: ◦ Distributed storage ◦ Fast access: network has high latency 2015-10-2536Qin Gao, LTI, CMU

37 Google Language Model Storage: ◦ Central storage or distributed storage How to deal with latency? ◦ Modify the decoder, collect a number of queries and send them in one time. It is a specific application, we still need something more general. 2015-10-2537Qin Gao, LTI, CMU

38 Again, made in Google: Bigtable It is the specially optimized for structured data Serving many applications now It is not a complete database Definition: ◦ A Bigtable is a sparse, distributed, persistent, multi-dimensional, sorted map 2015-10-2538Qin Gao, LTI, CMU

39 Data model in Bigtable Four dimension table: ◦ Row ◦ Column family ◦ Column ◦ Timestamp Row Column familyColumn Timestamp 2015-10-2539Qin Gao, LTI, CMU

40 Distributed storage unit : Tablet A tablet consists a range of rows Tablets can be stored in different nodes, and served by different servers Concurrent reading multiple rows can be fast 2015-10-2540Qin Gao, LTI, CMU

41 Random access unit : Column family Each tablet is a string-to-string map (Though not mentioned, the API shows that: ) In the level of column family, the index is loaded into memory so fast random access is possible Column family should be fixed 2015-10-2541Qin Gao, LTI, CMU

42 Tables inside table: Column and Timestamp Column can be any arbitrary string value Timestamp is an integer Value is byte array Actually it is a table of tables 2015-10-2542Qin Gao, LTI, CMU

43 Performance Number of 1000-byte values read/write per second. What is shocking: ◦ Effective IO for random read (from GFS) is more than 100 MB/second ◦ Effective IO for random read from memory is more than 3 GB/second 2015-10-2543Qin Gao, LTI, CMU

44 An example : Phrase Table Row: First bigram/trigram of the source phrase Column Family: Length of source phrase or some hashed number of remaining part of source phrase Column: Remaining part of the source phrase Value: All the phrase pairs of the source phrase 2015-10-2544Qin Gao, LTI, CMU

45 Benefit Different source phrase comes from different servers The load is balanced and the reading can be concurrent and much faster. Filtering the phrase table before decoding becomes much more efficient. 2015-10-2545Qin Gao, LTI, CMU

46 Another Example: GIZA++ Lexicon table: ◦ Row: Source word id ◦ Column Family: nothing ◦ Column: Target word id ◦ Value: The probability value With a simple local cache, the table loading can be extremely efficient comparing to current implemenetation 2015-10-2546Qin Gao, LTI, CMU

47 Conclusion Strangely, the talk is all about how Google does it A useful framework for distributed MT systems require three components: ◦ MapReduce software ◦ Distributed streaming data storage system ◦ Distributed structured data storage system 2015-10-2547Qin Gao, LTI, CMU

48 Open Source Alternatives MapReduce Library  Hadoop GoogleFS  Hadoop FS (HDFS) BigTable  HyperTable 2015-10-2548Qin Gao, LTI, CMU

49 THANK YOU! 2015-10-2549Qin Gao, LTI, CMU


Download ppt "Large Scale Machine Translation Architectures Qin Gao."

Similar presentations


Ads by Google