difficulty with semantic nets

2
STMIK TEKNOKRAT Bandarlampung Sistem Pakar kelas TI Gab Reg 1 Feri Saputra - 12312431 Nama : Feri Saputra NPM : 12312431 Kelas : TI Gab Reg 1 2.7 DIFFICULTIES WITH SEMANTIC NETS Although semantic nets can be very useful in representing knowledge, they have limitations, such as the lack of link name standards discussed previously. This makes it difficult to understand what the net is really designed for and whether it was designed in a consistent manner. A complementary problem to naming links is the naming of nodes. If a node is labeled “chair,” does it represent A specific chair The class of all chairs The concept of a chair The person who is the chair of a meeting Or something else? For a semantic net to represent definitive knowledge --- that is, knowledge that can be defined --- the link and node names must be rigorously defined. Of course, the same problems may occur in programming languages. Another problem is the combinatorial explosion of searching nodes, especially if the response to a query is negative. For a query to produce a negative result, many or all of the links in a net may have to be searched. As shown in the travelling salesman problem in Chapter 1, the number of links is the factorial of the number of nodes minus one if they are all connected. Although not all representations will require this degree of connectivity, the possibility of a combinatorial explosion exists. Semantic nets were originally proposed as models of human associative memory in which one node has links to others and information retrieval occurs due to a spreading activation of nodes. However, other mechanisms must also be available to the human brain for the reason that it does not take a long time for a 10 10 neurons in the human brain and about 10 15 links. If all knowledge was represented by a semantic net, it would take a very, very long time to answer negative queries like the football question because of all the searching involved with 10 15 links. Semantic nets are logically inadequate because they cannot define knowledge in the way that logic can. A logic representation can specify a certain chair, some chairs, all chairs, no chairs, and so forth, as will be discussed later in this chapter. Another problem is that semantic net are heuristically inadequate because there is no way to embed heuristic information in the net on how to efficiently search the net. A heuristic is a rule of thumb that may help in finding a solution but that is not guaranteed in the way an algorithm is guaranteed to find a solution. Heuristics are very important in AI because typical AI problems are so hard that an algorihmic solution does not exist or is too inefficient for practical use. The only standard control strategy built into a net that might help is inheritance, but not all problems may have this structure. A number of approaches have been tried to correct the inherent problems of semantic nets. Logic enhancements have been made and heuristic enhancements have been tried by attaching procedures to nodes. The procedures will be executed when the node becomes activated. However, the resulting systems gained little in conclusion of all this effort is that like any tool, semantic nets should be used for those things they do best, showing binary relationships, and not be distorted into a universal tool.

Upload: fery-saputra

Post on 09-Nov-2015

3 views

Category:

Documents


0 download

DESCRIPTION

Difficulty With Semantic Nets

TRANSCRIPT

  • STMIK TEKNOKRAT Bandarlampung Sistem Pakar kelas TI Gab Reg 1

    Feri Saputra - 12312431

    Nama : Feri Saputra

    NPM : 12312431

    Kelas : TI Gab Reg 1

    2.7 DIFFICULTIES WITH SEMANTIC NETS

    Although semantic nets can be very useful in representing knowledge, they have limitations,

    such as the lack of link name standards discussed previously. This makes it difficult to understand what

    the net is really designed for and whether it was designed in a consistent manner. A complementary

    problem to naming links is the naming of nodes. If a node is labeled chair, does it represent

    A specific chair

    The class of all chairs

    The concept of a chair

    The person who is the chair of a meeting

    Or something else? For a semantic net to represent definitive knowledge --- that is, knowledge

    that can be defined --- the link and node names must be rigorously defined. Of course, the same problems

    may occur in programming languages.

    Another problem is the combinatorial explosion of searching nodes, especially if the response

    to a query is negative. For a query to produce a negative result, many or all of the links in a net may

    have to be searched. As shown in the travelling salesman problem in Chapter 1, the number of links is

    the factorial of the number of nodes minus one if they are all connected. Although not all representations

    will require this degree of connectivity, the possibility of a combinatorial explosion exists.

    Semantic nets were originally proposed as models of human associative memory in which one

    node has links to others and information retrieval occurs due to a spreading activation of nodes.

    However, other mechanisms must also be available to the human brain for the reason that it does not

    take a long time for a 1010 neurons in the human brain and about 1015 links. If all knowledge was

    represented by a semantic net, it would take a very, very long time to answer negative queries like the

    football question because of all the searching involved with 1015 links.

    Semantic nets are logically inadequate because they cannot define knowledge in the way that

    logic can. A logic representation can specify a certain chair, some chairs, all chairs, no chairs, and so

    forth, as will be discussed later in this chapter. Another problem is that semantic net are heuristically

    inadequate because there is no way to embed heuristic information in the net on how to efficiently search

    the net. A heuristic is a rule of thumb that may help in finding a solution but that is not guaranteed in

    the way an algorithm is guaranteed to find a solution. Heuristics are very important in AI because typical

    AI problems are so hard that an algorihmic solution does not exist or is too inefficient for practical use.

    The only standard control strategy built into a net that might help is inheritance, but not all problems

    may have this structure.

    A number of approaches have been tried to correct the inherent problems of semantic nets. Logic

    enhancements have been made and heuristic enhancements have been tried by attaching procedures to

    nodes. The procedures will be executed when the node becomes activated. However, the resulting

    systems gained little in conclusion of all this effort is that like any tool, semantic nets should be used for

    those things they do best, showing binary relationships, and not be distorted into a universal tool.

  • STMIK TEKNOKRAT Bandarlampung Sistem Pakar kelas TI Gab Reg 1

    Feri Saputra - 12312431

    2.7 Kesulitan dengan jaringan semantik

    Meskipun jaringan semantik dapat sangat berguna dalam menunjukkan pengetahuan, tetapi

    jaringan tersebut mempunyai batasan seperti kekurangan standard nama link yang didiskusikan

    sebelumnya. Hal ini akan membuatnya sukar untuk memahami apa sebenarynay jaringan yang di desiain

    untuk dan apakah didesain dengan cara yang konsisten. Problem komplementary pada link pemberian

    nama adalah pembuatan nama node. Jika suatu node diberi label dengan Chair, maka menunjukkan :

    A specific chair

    The class of all chairs

    The concept of a chair

    The person who is the chair of a meeting

    Atau arti lain? Untuk jaringan semantik dapat menunjukkan pengetahuan definite, yaitu, pengetahuan yang dapat ditentukan, link dan nama node harus secara kuat ditentukan. Tentu saja,

    problem yang sama mungkin terjadi dalam bahasa pemrograman.

    Problem lain adalah eksplosi kombinasi dari node penelitian, khususnya jika yang merespon ke

    query adalah negatif. Yaitu, untuk query yang membuat hasil negatif, beberapa tau seluruh link dalam

    jaringan harus diteliti. Seperti ditunjukkan dalam problem pengiriman/perjalanan salsesman dalam bab-

    01, sejumlah link merupakan faktorial dari sejumlah node minus satu jika seluruhnya dihubungkan.

    Meskipun tidak semua representasi akan memerlukan tingkat hubungan ini, namun kemungkinan dari

    eksplosi kombinasi akan muncul.

    Jaringan semantik, aslinya diusulkan sebagai model memory gabungan manusia dimana satu

    node mempunyai link ke yang lainnya dan penerimaan informasi terjadi karena penyebaran aktifasi

    node. Namun demikian, mekanisme lain harus juga ada pada pikiran manusia sejak tidak memerlukan

    waktu lama bagi manusia untuk menjawab pertanyaan adakah team sepak bola di pluto? Ada sekitar 10 pangkat 10 neuron dalam pikiran manusa dan kira-kira 10 pangkat 15 link. Jika semua pengetahuan

    ditunjukkan dengan jaringan semantik,maka akan memerlukan waktu yang sangat lama untuk menjawab

    pertanyaan negatif seperti pertanyaan sepak bola karena seluruh penelitian dicakup dalam 10 pangkat

    15 link.

    Jaringan semantik secara logikal tidak memadai karena tidak dapat menentukan pengetahuan

    dengan cara yang dapat dilakukan oleh logika.

    Representasi logika dapat menentukan kursi tertentu, beberapa kursi, seluruh kursi, tak ada kursi, dan

    sebagainya seperti yang akan didiskusikan kemudian dalam bab ini. Problem lain adalah bahwa jaringan

    semantik secara heuristik tidak memadai karena tidak ada cara untuk memancangkan informasi heuristik

    dalam jaringan tas bagaimana mengefisiensi penelitian jaringan. Heuristic merupakan baris (thumb) yang mungkin membantu dalam menemukan solusi tetapi tidak dijamin seperti algoritma yang

    menjamin solusi. Heuristic sangatlah penting dalam AI karena problem AI tipikal begitu sukar/keras

    dimana solusi algoritma tidak akan muncul atau terlalu tidak efisien untuk penggunaan praktis. Satu-

    satunya strategi kontrol standard dibuat ke dalam jaringan yang mungkin membantu adalah pewarisan

    tetapi, tidak semua problem mempunyai struktur ini.

    Sejumlah pendekatan telah dicobakan untuk membenarkan problem pewarisan dari jaringan

    semantik. Penambahan logika telah dibuat, dan penambahan heuristic telah dicobakan dengan

    melawankan prosedur pada node. Prosedur akan dibuat jika node menjadi aktif. Namun demikian,

    system yang dihasilkan diperoleh kecil dalam kemampuan pada biaya jaringan semantik yang dapat

    diekspresikan natural. Kesimpulan dari seluruh usaha ini adalah bahwa seperti suatu peralatan, jaringan

    semantik harus digunakan untuk seseuatu yang mereka kerjakan terbaik, dengan menunjukkan

    hubungan binary, dan tidak disimpan ke dalam peralatan universal.